Short-Term Effects of Cerebellar tDCS on Standing Balance Performance in Patients with Chronic Stroke and Healthy Age-Matched Elderly

Subjects

Protocol

All subjects first participated in an intake and clinical assessment session, after which they returned for respectively two (healthy controls) or three (patients) sessions, in which they had to perform a medio-lateral postural tracking task while being stimulated with anodal cb_tDCS. Directly before and after this tracking task with cb_tDCS, standing balance performance was assessed during several quiet standing tests. These sessions were minimally one and maximally two weeks apart. For a schematic overview of the protocol, see Fig. 1.

Open image in new window

Tracking Task

Two dots were presented on a video screen located at eye height for the tracking task (Fig. 2b). One dot represented a moving target, which the subject was asked to follow as precisely as possible with the second tracking dot, representing their CoP measured with a force plate. The CoP signal was low pass filtered with a second-order Butterworth filter at 10 Hz (D-flow, Balance Workstation, Motek, Amsterdam, The Netherlands). Four repetitions of the tracking task of 3 min each were performed by each subject. In two of those repetitions, the target moved in a predictable manner with an increasing velocity in eight blocks of 20 s from 0.16 to 1.28 cm/s. During the two other repetitions, the target moved in a pseudo-random manner with 16 blocks of 10 s with a velocity between 0.16 and 1.28 cm/s. Those repetitions were therefore considered to be unpredictable in terms of velocity. Subjects received feedback on their performance after each repetition. Performance was defined as the percentage of the time of accurate overlap of the target and tracking dot during the eight different velocities. Subsequently, subjects could rest as long as needed between repetitions.

Open image in new window

Performance on the Tracking Task

To evaluate a possible learning effect of the tracking task, performances of the last two repetitions were averaged over all frequencies. This included one repetition with a predictable and one with an unpredictable moving target.

tDCS

During the tracking task, cb_tDCS was delivered by a battery-driven portable stimulator (Starstim ®; Neuroelectrics, Barcelona, Spain) through 3.14-cm² electrodes filled with conducting gel, with a direct 1.5 mA current. The anodal electrode was placed 3 cm lateral of the inion. Two cathodal electrodes were placed on the ipsi-lateral buccinators muscles (Fig. 2a). Anodal stimulation was applied for 20 min with 1.5 mA in all sessions. During the first 3 min of stimulation, subjects were sitting on a chair to get acquainted with the stimulation while the tracking task was explained. For patients, three experimental sessions, with minimally one week and maximally two weeks apart, were performed: (1) anodal cb_tDCS on contra-lesional side, (2) anodal cb_tDCS on the ipsi-lesional side, and (3) sham cb_tDCS on the contra-lesional side. In sham tDCS, the current was automatically switched off after 30 s and switched on again for the last 30 s of the 20-min stimulation period. Healthy controls only had two sessions, both ipsi-lateral to their dominant leg, with either anodal or sham stimulation. The order of the sessions was randomized. The stimulation was always finished before subjects finished their fourth repetition of the tracking task. The dominant leg was determined by the preferred leg used and better performance on the lateralized items of the BBS (i.e., looking over the shoulder, turning, tandem stance, and one-leg stance).

Standing Balance Performance Assessment

Before and after the stimulation, standing balance performance was assessed during several quiet standing positions. Ground reaction forces were measured during these positions with a sample frequency of 1000 Hz using an 80 × 80 cm force plate (Motek, Amsterdam, The Netherlands). The analog signals were converted with a 16-bit analog-to-digital converter with a 10-V range (National Instruments, Austin, USA).

Subjects were asked to stand quietly in three sequential standing positions: (1) with eyes open, (2) eyes closed, and (3) in the most challenging, subject-specific (semi-) tandem stance, on a force plate during five repetitions of 1 min. Subjects rested for a minimum of 30 s in between standing tests. The rest period was extended upon the subjects’ request to avoid fatigue. In standing positions 1 and 2, subjects stood barefoot with their arms relaxed, alongside the trunk if possible, with their feet at hip width in 9° exorotation (Fig. 2c). During the tandem stance, subjects were asked to stand in the most difficult position which they could hold for 1 min as determined in the first clinical assessment session. The foot positions of the first session were marked and measured to keep foot placement the same in the consecutive sessions.

At four time points during a session, subjects were asked to indicate their level of headache, nausea, fatigue, and depressed mood on a visual analog scale (VAS) from 0 to 100 mm, in order to track commonly reported side effects of tDCS [29].

All subjects were told at the end of all measurements that during one of the sessions, a placebo stimulation was given and asked if they could indicate which session it was.

Clinical Assessments

The clinical measures were performed by the researcher according to recommended guidelines [2] before the sessions with tDCS. The following assessments were performed: Berg Balance scale (BBS) [30], functional reach task (FRT) [31], timed up and go (TUG) [32], fall efficacy scale (FES) [33], fall history, and Erasmus modification of the Nottingham Sensory Assessment of the Lower Extremity (EmNSA-LE) [34]. For healthy controls, the dominant leg was determined by the preferred leg used and better performance on the lateralized items of the BBS (i.e., looking over the shoulder, turning, tandem stance, and one-leg stance). The most difficult position for subjects to hold for more than 30 s was determined with an ordinal scale ranging from (1) full tandem stance with non-paretic leg/dominant in front, (2) non-paretic/dominant leg a step ahead, (3) non-paretic/dominant leg half a step ahead, (5) full tandem stance with paretic leg/non-dominant in front, (6) paretic/non-dominant leg a step ahead, (7) paretic/non-dominant leg half a step ahead, or (8) feet placed together.

Stroke patients were also assessed with the Fugl–Meyer Motor Assessment of the Lower Extremity (FMA-LE) [35], Motricity Index (MI) [36], Nottingham Extended Activities of Daily Living index (NEADL) [37], and O-Letter Cancelation test (LCT) [38].

Data Analysis and Pre-Processing

Recorded data were processed using MATLAB 2012a (the MathWorks, Inc., Natick, MA, USA). Statistical analysis was performed using IBM SPSS statistics version 22 (IBM Corporation, Armonk, NY, USA).

Outcome Parameters

Force plate data were low pass filtered with a second-order Butterworth filter with a cutoff frequency of 10 Hz, after which the CoP in both the anterior–posterior (AP) and the medio-lateral (ML) direction was calculated. The middle 50 s of each trial were used for further analysis. The signals were linear detrended with a period of 20 s. The following parameters were computed from the sum of the vectors of the CoP in the AP and ML direction: the mean amplitude of the CoP (ACoP) calculated as the root mean squared distance from the mean CoP, its standard deviation represents the amplitude’s variability (varCoP), the velocity of the CoP (VCoP) calculated by the sum of the distance between sequential points divided by its length, its standard deviation represents the velocity’s variability (varVCoP), and the range determined as the maximal difference between any two points of the time series. A CoP composite score of the abovementioned parameters was calculated as a comprehensive outcome parameter representing qualitative aspects of standing balance performance [39]. Each parameter, calculated for ML and AP direction separately, was transformed to a z-score calculated over all parameters, separately for the stroke patients and the healthy controls. This transformation leads to a mean of 0 with a standard deviation of 1, making it possible to average the ten transformed parameters per standing position into a CoP composite-score (comp-score).

Statistical Analysis

Normality was checked by visual inspection of the probability distribution (q–q plot) and the box plot. A Shapiro–Wilks test was performed on the data or the residual when appropriate. When the assumptions of normality were not met, a natural log transformation was applied after which normality was checked again. In case this transformation was not sufficient, or in case of ordinal or nominal data, the non-parametric equivalent of the below mentioned tests was used. The significance level α was set two-tailed at 0.05.

CoP baseline differences between healthy controls and patients were analyzed with an independent t test with a Bonferroni correction for multiple testing per standing position. To correct for differences in sample size, Hedges’ g (g) was used to calculate effect sizes, when significant differences were found.

A generalized estimating equation (GEE) model, with a correction for baseline CoP comp-score, was used to establish an association between the stimulation conditions on CoP comp-scores for patients and healthy control separately. The model was also tested for confounding of randomization order. If the β values for the stimulation conditions changed with more than 10%, this was considered an improvement of the model. In case of a significant association between post-measurement CoP comp-score and cb_tDCS, the parameters from which the CoP comp-score was constructed were evaluated separately.

Performance on the Tracking Task

A GEE model was used to evaluate the improvement of performance on the tracking task over time. Stimulation condition was added as a possible confounder to examine if performance was influenced by cb_tDCS.

Responders and Non-Responders

To investigate the differences in response to cb_tDCS between subjects (interindividual response), a distinction was made between patients who showed a response on the CoP comp-score and patients who did not. A subject was defined as a “responder” when a change in CoP comp-score (post–pre) for the tandem stance was more than one standard deviation larger in the contra-lesional stimulation or the ipsi-lesional condition compared to the sham condition. Differences in baseline CoP comp-score, fatigue, and clinical and demographic characteristics between responders and non-responders were analyzed with independent t tests.

Procedures

All subjects completed the experimental protocol. One patient was not able to perform five repetitions of each static assessment of standing balance performance due to fatigue. Three repetitions of each assessment were performed instead.

Two patients were able to perform a semi-tandem stance for 60 s with the non-paretic leg ahead of the paretic leg, ten patients were able to perform a semi-tandem stance with the non-paretic leg a small step in front of the paretic leg, and three patients were able to perform a semi-tandem stance with the paretic leg a small step in front of the non-paretic leg. Nine healthy controls were able to perform a tandem stance with the dominant leg ahead, and one healthy control performed a semi-tandem stance with the dominant leg ahead.

Successfulness of Blinding Procedures

When asked to indicate the sham condition at the end of the protocol, four out of ten healthy controls answered correctly, five incorrectly, and one could not make a choice. From the patients, seven answered correctly, six incorrectly, and two patients could not make a choice.

Possible Side Effects, Fatigue

No subjects reported headaches or nausea during any of the sessions. Three healthy controls reported to be somewhat fatigued, but no differences in VAS on fatigue between sham: median (md) = 0, interquartile ranges (iqr) = 0–0.15 and cb_tDCS: md = 0, iqr = 0–0, were found, z = − 0.54, p = 0.60. Eleven patients reported a higher VAS on fatigue after stimulation, but no differences were found between the sham: md = 1.1, iqr = 0–3; contra-lesional: md = 1, iqr = 0–2.6; and ipsi-lesional: md = 1.3, iqr = 0–2.1, χ² = 0.98, p = 0.61 conditions. Only one patient reported 7 out of 100 points on the VAS for depressed mood at the start of one of the sessions; no increase was reported after stimulation.

Baseline Differences Between Patients and Healthy Controls

The baseline CoP parameters are displayed in Fig. 3. Patients showed significant larger excursion on all baseline CoP parameters for the eyes open; ACoP: mean healthy = 2.90 mm, sd = 0.47, mean patients = 4.17 mm, sd = 1.53, CI of the mean difference = − 2.16 to − 0.38, t(17.6) = − 3.01, g = 1.00; varCoP: healthy = 3.67 mm, sd = 0.58, patients = 5.35 mm, sd = 1.99, CI = − 2.83 to − 0.53, t(17.3) = − 3.07, g = 1.02; range: healthy = 20.68 mm, sd = 0.22, patients = 32.21 mm, sd = 0.42, CI = − 18.88 to − 4.17, t(16.4) = − 3.31, g = 31.37; VCoP: md healthy = 7.10 mm/s, iqr = 5.88–7.81, md patients = 10.78 mm/s, iqr = 8.80–15.48, CI = 0.43–0.78, ratio due to log transformation, t (23) = 0.02, g = 1.71; varVCoP: md healthy = 5.79 mm/s, iqr = 5.79–6.62, patients = 8.81 mm/s, iqr = 7.76–13.87, CI = 0.40 to 0.75, ratio due to log transformation, t(23)=0.02, g = 1.57, all p < 0.05.

Open image in new window

This was also the case for the eyes closed position when compared to healthy age-matched controls; ACoP: mean healthy = 3.66 mm, sd = 0.65, mean patients = 5.71 mm, sd = 2.02, CI = − 3.23 to − 0.88, t(18) = − 3.67, g = 1.22; varCoP: healthy = 4.66 mm, sd = 0.83, patients = 7.21 mm, sd = 2.57, CI = − 4.06 to − 1.07, t (18)= − 3.60, g = 1.20, range: healthy = 26.49 mm, sd = 5.23, patients = 42.16 mm, sd = 15.20, CI = − 24.59 to − 6.74, t(18.5) = − 3.68, g = 1.23; VCoP: healthy = 11.62 mm/s, sd = 2.28, patients = 20.44 mm/s, sd = 8.91, CI = − 13.91 to − 3.73, t(16.6) = −3.66, g = 1.20; varVCoP: healthy = 9.42 mm/s, sd = 1.94, patients = 17.26 mm/s, sd = 7.56, CI = − 12.16 to − 3.51, t(16.6) = − 3.83, g = 1.26, all p < 0.05. For the (semi-)tandem stance position, no significant differences were found between patients and healthy controls; ACoP: md healthy = 5.17 mm, iqr = 4.65–6.34, md patients = 5.86 mm, iqr = 4.65–7.89, ratio due to log transformation, CI = 0.63 to 1.15, t (23) = 0.33; varCoP: md healthy = 6.55 mm, iqr = 5.80–8.45, md patients = 7.31 mm, iqr = 5.82–10.09, ratio due to log transformation, CI = 0.63 to 1.15, t[23]=0.33; range: md healthy = 40.29 mm, iqr = 36.65–55.94, md patients = 47.72 mm, iqr = 33.14–64.73, ratio due to log transformation, CI = 0.64 to 1.17, t(23) = 0.37; md VCoP: healthy 30.25 = mm/s, iqr = 26.22–32.48, md patients = 26.33 mm/s, iqr = 23.90–36.83, U = 61, z = − 0.78; varVCoP: md healthy 26.24 = mm/s, iqr = 21.40–28.96, md patients = 23.84 mm/s iqr = 19.47–31.16, U = 71, z = − 0.22, all p > 0.5.

Effect of Stimulation on CoP Parameters

The tested model revealed no significant changes in CoP comp-score associated with cb_tDCS in the eyes open: β = 0.02, CI = − 0.09 to 0.12, p = 0.73; eyes closed: β = 0.08, CI = − 0.01 to 0.16, p = 0.07; and tandem: β = − 0.08, CI = − 0.41–0.25, p = 0.64 for the healthy controls, see Table 2. Adding the stimulation order to the model did not change β values with more than 10%.

In the patient group, a significant association between contra-lesional stimulation and a decrease in CoP comp-score in the tandem position was found: β = − 0.25, CI = − 0.48 to − 0.03, p = 0.03. Post hoc analysis showed a significant decrease in ACoP: β = − 0.86, CI = − 1.58 to − 0.15, p = 0.02; varCoP: β = − 1.10, CI = − 1.93 to − 0.26, p = 0.01; range: ratio due to log transformation, β = 0.94, CI = 0.90–0.98, p = 0.01; and VCoP: ratio due to log transformation, β = 0.97, CI = 0.94 to 0.99, p = 0.02 but not in varVCoP: ratio due to log transformation, β = 0.97, CI = − 0.93 to 1.01, p = 0.11, see Fig. 4. The GEE-model constructed for the eyes open position revealed a significant association between ipsi-lesional stimulation and a lower CoP comp-score: β = − 0.09, CI = − 0.18 to − 0.01, p = 0.03; after correcting for randomization order, the association was no longer significant and changed to β = 0.00, CI = − 0.09 to 0.90, p = 0.94.

Open image in new window

No changes in CoP comp-score in the eyes closed position associated with cb_tDCS were found for patients, contra-lesional stimulation: β = 0.02, CI = − 0.11–0.16, p = 0.73 and ipsi-lesional stimulation: β = − 0.01, CI = − 0.19–0.16, p = 0.89. Adding the stimulation order to the model changed β values with more than 10% to contra-lesional stimulation: β = 0.06, CI = − 0.10–0.22, p = 0.44 and ipsi-lesional stimulation: β = 0.09, CI = − 0.08–0.26, p = 0.30 (Fig. 5). See Table 2 for an overview of the corresponding β values and confidence intervals.

Open image in new window

Performance on the Tracking Task

The healthy controls showed a significantly higher tracking task performance during the second measurement, m = 99.33, sd = 0.57, β = 0.80, CI = 0.10 to 1.50, p = 0.03, as compared to the first measurement, m = 98.53, sd = 1.31. No effect of stimulation was found β = 0, CI = − 0.70 to 0.70, p = 1.0.

The patient group showed a significantly higher tracking task performance during the second measurement; m = 92.90, sd = 7.5, with a ratio of β = 1.62, CI = 1.23 to 2.14, p < 0.01 and on the third measurement, m = 94.60, sd = 6.11, with a ratio of β = 2.06, CI = 1.51 to 2.80, p < 0.01 as compared to the first measurement, m = 88.60, sd = 9.29.

No effect of contra-lesional stimulation was found with a ratio due to log transformation of β = 1.15, CI = 0.89 to 1.50, p = 0.29, nor an effect of ipsi-lesional stimulation, β = 1.16, CI = 0.91–1.49, p = 0.23.

Responders and Non-responders

Baseline Differences Between Patients and Healthy Controls

The increase in postural sway found in stroke patients compared to aged-matched controls and the enlarged sway with the more difficult positions are in line with previously found results in a comparable population (46). However, no significant differences in CoP parameters were found between stroke patients and controls for the tandem stance position. This result may be explained by the individualized feet positioning during the tandem stance per subject. None of the stroke patients could hold the full tandem stance position with either leg behind, while all but one healthy subject could hold this position with the non-preferred leg behind. This again indicates that the task may not have been difficult enough for the healthy controls. The main purpose of this proof-of-concept study was to investigate the potential of cb_tDCS to elicit qualitative changes in standing balance performance in patients with a stroke. Since the differences between patients are much larger than the effect that can be expected from a single training session, a within-subject design was needed with a challenging task, tailored to the specific capacity of each patient.

Outcome Parameters

A decrease in CoP parameters is generally assumed to reflect an improvement in postural stability in patients with a stroke (47). There is a strong need for more sensitive and reliable measures to be able to quantify subtle changes in standing balance performance and disentangle postural control mechanisms (48). Despite several promising methods to quantify postural control, a golden standard is still lacking (49, 50). We used the CoP comp-score as a sensitive comprehensive outcome parameter, combining information from five parameters of standing balance performance (39). A sensitive outcome parameter, able to detect subtle qualitative changes, is also needed to establish an optimal dose in terms of sessions and intensity. To relate the currently found short-term effects of anodal cb_tDCS on standing balance performance to a clinical meaningful and long-term improvement, the effect of multiple training sessions should be measured using both qualitative parameters of standing balance performance and clinical outcome measures.

Responders and Non-Responders

A general point of concern in tDCS research is the different responsiveness of subjects to the stimulation, i.e., why do some subjects respond to the stimulation while others do not? Several reviews on both healthy subjects as well as patients with stroke have highlighted factors such as age, anxiety, time since stroke, lesion type, lesion location, and motor function as contributors to interindividual variability in response to tDCS (51, 52, 53, 54, 55). Within this study, ten patients showed changes toward normal CoP values in response to contra-lesional cb_tDCS compared to sham. On a group level, this contrast was large enough for a significant association; the non-responders should, however, not be ignored. These interindividual differences play a role in many tDCS studies, which lead to ambiguity in the interpretation of results and the conclusion of some authors that tDCS does not have any added value or vice versa (56). Within the small sample of this study, no differences could be detected between responders and non-responders on several subject characteristics. Next to clinical characteristics, the initial state of neuronal populations could play a role in terms of responsiveness to the stimulation (54, 57). Neuroimaging techniques could provide valuable insights into these neuronal state dependencies (58).

Strength and Limitations

This proof-of-concept study is the first aimed at the short-term effect on standing balance performance of a single training sessions combined with anodal cb_tDCS in stroke patients. The current setup was able to detect subtle qualitative changes in standing balance performance by using high-resolution kinetic parameters. The study population was however small and heterogeneity in terms of lesion type, location, and motor function could have influenced the interindividual variability in response to the stimulation. Within the sample, two patients were included with a lesion primarily located in the brainstem; both of these patients were classified as responders to the contra-lesional cb_tDCS. The current sample is, however, too small to perform a sub-group analysis or to generalize these findings to a wider population. The healthy controls in this study only received stimulation ipsi-lateral to their dominant leg. It is possible that an improvement in the healthy subjects could have been found after contralateral stimulation. The more evident explanation for the lack of improvement in healthy controls is, however, that the task was too easy and a more challenging task should have been chosen for this group.

Future Research

The current study shows the potential of anodal cb_tDCS for improving standing balance performance in a chronic stroke population, though interindividual differences should be further studied. Analysis of ongoing cortical processes during standing balance tasks and the influence of cb_tDCS on these processes may give more insights in unknown underlying mechanisms. Moreover, these unknown mechanisms leading to interindividual differences could be very different in the more acute phase after stroke. Behavioral restitution of function, mainly taking place in the first 8–12 weeks post-stroke (59), go alongside changes in growth factors, creating a critical time window for recovery (60, 61, 62). tDCS might be able to optimize the learning potential in this time window, leading to a completely different paradigm of the mechanisms of action of tDCS in the sub-acute phase than in a population of patients in a chronic phase after stroke (63, 64). Considering that balance recovery is especially important in the early phase after stroke, tDCS interventions should also be tested in this time window of enhanced neuroplasticity. In addition, possible contributors to interindividual differences, such as a wide variety of clinical characteristics as well as neuronal state, should be recorded.