The “broken escalator” phenomenon: Vestibular dizziness interferes with locomotor adaptation

2.1 Participants

Twenty patients with unilateral vestibular neuritis (5 female; mean age 54.8 years, SD = 14.4), fifteen patients with unilateral Meniere’s disease (5 female; mean age 48.9 years, SD = 12.3) and fifteen healthy controls with normal vestibular function and no history of vertigo (6 female; mean age 55 years, SD = 7.5) were recruited. Participants were age-matched (independent samples t-test P > 0.096).

Meniere’s disease was diagnosed according to American Academy guidelines [1] and most patients conducted this experiment before participating in a trial of intratympanic injection for unilateral refractory Meniere’s disease [39]. All vestibular neuritis patients had a typical history of sudden onset rotational vertigo lasting for several days, spontaneous unidirectional nystagmus, a positive head-impulse test, normal hearing and a clinically significant (>25%) unilateral caloric paresis, as previously described [11]. Vertigo onset took place at least six months previously, with no further attacks.

2.2 Equipment

The motion stimulus was provided by a linear sled running on a level track, powered by two linear induction motors [42] controlled by sled velocity as recorded with a tachometer. Anterior-posterior trunk position was measured using a Fastrak™ system (Polhemus, VT, USA) that sampled at 250 Hz. The movement sensor was secured at the level of the C7 vertebra to measure linear trunk displacement. Step timing was measured by contact plates under each foot and corroborated with a sled-mounted linear accelerometer.

2.3 Experimental design

2.3.1 Locomotor experiment

The experimental sequence involved three stages performed in this order: BEFORE (5 trials, stationary sled), MOVING (5 trials, moving sled) and AFTER trials (5 trials, stationary sled) [8].

In BEFORE, MOVING and AFTER trials, participants stepped from a stationary platform onto the sled, see Fig. 1. Participants began by facing the direction of movement and initiated a step (right foot first) from a stationary stance, prompted by a single auditory beep, and continued their walk on to the sled. Participants were instructed that once both feet were on the sled they should stop and remain still [8].

OPEN IN VIEWER

Fig. 1

Experimental design. The figure shows the experimental sequence (from left to right) and model trunk data from a Fastrak sensor placed on C7. All groups performed BEFORE, MOVING and AFTER trials in which subjects step, leading with the right leg, from a fixed surface to a sled which is either stationary or moving. Forward sway was measured in stationary sled phases (BEFORE and AFTER trials) as the maximum forwards deviation of the trunk relative to the final 3 s of the trial. Trunk sway in the MOVING phase was measured as the peak-to-peak forwards movement. In all phases (BEFORE, MOVING and AFTER) approach velocity was measured in a 0.5 s epoch prior to foot-sled contact. In the MOVING phase, stepping onto the moving sled for the first time results in backwards sway. A significant locomotor aftereffect is characterised by a forward trunk sway overshoot (12 cm in the representative trace in the figure) and faster gait velocity in AFTER trials, despite explicit information that the sled will not move.

In MOVING trials, the onset of sled motion was triggered by breaking an infra-red light beam when the subject stepped forward from the start position. This resulted in the sled moving, after a 600 ms delay, a distance of 3.7 m in 4.2 s; (maximum velocity 1.4 m/s achieved at 1.3 s). Participants were asked to avoid using the handrails unless they truly felt they could fall. Upon completion of the MOVING trials, participants performed the AFTER trials. They were given the following instruction: “I want you to step onto the sled as before. Only this time it is not going to move” –and the motor was ostensibly turned off for reassurance. Each trial lasted 16 seconds after which the participant returned to the original starting position.

3.1 Locomotor adaptation (MOVING trials)

MOVING trials, in which participants were asked to step onto a moving sled, were used to compare adaptation between groups (i.e., the level of postural sway and gait velocity between successive trials). As shown in Fig. 2, trunk sway was larger in patients compared to controls (Fig. 2). All groups had a reduction of trunk sway with trial number demonstrating adaptation (significant trial number effect in two-way Mixed ANOVA, below). However, the rate of adaptation was quicker in controls compared to vestibular neuritis and Meniere’s disease patients, as shown by a steeper adaptation gradient in MOVING trials 1–3, where maximal learning occurs (independent samples t-tests: vestibular neuritis vs Control, P = 0.035); Meniere’s disease vs Control, P = 0.033).

OPEN IN VIEWER

Fig. 2

Locomotor performance of control and patient groups. Mean±Standard Error (SEM) A. Vestibular neuritis vs. controls and B. Meniere’s disease vs. controls during BEFORE, MOVING and AFTER phases for trunk sway (top row) and gait velocity (bottom row). The horizontal axis shows the trial number (1–5). The trunk sway aftereffect (AFTER trial 1) was larger in Meniere’s disease compared to controls. Gait velocity was slower in patients compared to controls in both the BEFORE and AFTER phases. Significant differences between VN and controls and MD and controls are shown with asterisks *P < 0.05, **P < 0.01, ***P < 0.001.

A correlation was performed on MOVING trial data for the degree of adaptation in patients. The rationale was that excessive body sway may interfere with adaptation to the moving task and, indeed, a reduced degree of adaptation in trials 1–3 was associated with higher average trunk sway in MOVING trials (average1–5) (r = 0.35, P = 0.041, Fig. 3).

OPEN IN VIEWER

Fig. 3

Association between the degree of adaptation (i.e. change in trunk sway between MOVING trials 1–3) and mean trunk sway across MOVING trials. The figure shows that subjects with larger levels of overall trunk sway have a slower rate of adaptation. Asterisk: P < 0.05.

3.2 Locomotor aftereffect (BEFORE vs. AFTER trials)

All groups showed an aftereffect, that is, they all produced a significantly larger trunk overshoot and faster gait velocity in the first AFTER trial compared to baseline (i.e., mean BEFORE trials 3–5, P < 0.001, paired samples t-test).

3.3 Dizziness, vertigo and canal paresis

As shown in Table 1, DHI (vestibular neuritis = 20.4, Meniere’s disease = 48.3) and VSS scores (vestibular neuritis = 10.1, Meniere’s disease = 20.8) were twice as high in Meniere’s disease compared to vestibular neuritis (both P < 0.001, independent samples t-test). Higher VSS scores were associated with higher DHI scores (vestibular neuritis: Pearson’s r = 0.64, P = 0.002; Meniere’s disease: Pearson’s r = 0.72, P = 0.002) indicating agreement between the clinical scales.

Caloric paresis (for vestibular neuritis = 33.5%, SD = 23.2; Meniere’s disease = 43.8%, SD = 25.2) was marginally larger in Meniere’s disease compared to vestibular neuritis (Table 1, P = 0.038, independent samples t-test), but there was no correlation between caloric paresis and DHI scores (vestibular neuritis: Pearson’s r = 0.11, P = 0.65; Meniere’s disease: Pearson’s r = 0.37, P = 0.17) or between caloric paresis and VSS scores in vestibular neuritis (Pearson’s r = 0.08, P = 0.78) in line with previous work [12, 38] or in Meniere’s disease. The correlation between caloric paresis and VSS scores approached significance in Meniere’s disease: (Pearson’s r = 0.35, P = 0.055).

Next, we investigated whether overall symptom load impacts on locomotor parameters. When there was a significant association, we isolated the effect using DHI (physical, functional and emotional) or VSS subscales (autonomic-anxiety and vertigo-imbalance) as appropriate.

3.3.1 Vestibular neuritis

11/21 vestibular neuritis patients were poorly compensated i.e., DHI score >20 [38]. On the other end of this spectrum 6 patients were virtually symptom free, including 4 who scored “0” on the DHI.

MOVING trials: Higher DHI (Pearson’s r = 0.40, P = 0.041, Fig. 4A) and VSS scores (Pearson’s r = 0.53, P = 0.016) were associated with larger mean trunk sway in MOVING trials, specifically higher functional DHI scores (Pearson’s r = 0.57, P = 0.009) with a trend relationship to higher autonomic-anxiety VSS scores (Pearson’s r = 0.36, P = 0.053). Mean trunk sway was not related to caloric paresis. The rate of adaptation and the degree of adaptation were not associated with DHI, VSS or caloric paresis. Mean gait velocity in MOVING trials was not associated with DHI, VSS scores or caloric paresis.

OPEN IN VIEWER

Fig. 4

Associations between mean trunk sway amplitude across MOVING trials and Dizziness Handicap Inventory (DHI) score for A. Vestibular neuritis and B. Meniere’s disease. Together, A and B show that the size of mean trunk sway correlates with the level of subjective dizziness as shown by the Pearson’s r correlation coefficient score. Asterisks display the level of significance, where *P < 0.05.

AFTER trial 1: A larger trunk aftereffect correlated to higher DHI scores (Pearson’s r = 0.53, P = 0.016, Fig. 5A), specifically higher functional DHI scores (Pearson’s r = 0.60, P = 0.005). The association between the trunk aftereffect and emotional DHI scores approached significance (Pearson’s r = 0.41, P = 0.06; Fig. 6). Deeper analysis showed a significant correlation between the trunk aftereffect and emotional DHI scores after the four asymptomatic (i.e., patients that did not experience emotional effects from their dizziness; DHI = 0) vestibular neuritis patients were omitted from analysis (Pearson’s r = 0.51, P = 0.041). There was no association between the size of the trunk aftereffect and caloric paresis. The gait velocity aftereffect was not associated with DHI, VSS scores or caloric paresis.

OPEN IN VIEWER

Fig. 5

Associations between trunk aftereffect size (AFTER trial 1) and Dizziness Handicap Inventory (DHI) score for A. Vestibular neuritis and B. Meniere’s disease. Together, A and B show that the size of the trunk aftereffect correlates with the level of subjective dizziness as shown by the Pearson’s r correlation coefficient score. Asterisks display the level of significance, where *P < 0.05.

OPEN IN VIEWER

Fig. 6

Association between aftereffect size and emotional Dizziness Handicap Inventory (DHI) scores for Meniere’s disease and vestibular neuritis patients. The correlation was significant in Meniere’s disease (Pearson’s r = 0.62, P = 0.012) but not for vestibular neuritis patients (Pearson’s r = 0.41, P = 0.06). However, when asymptomatic vestibular neuritis patients were omitted (i.e. score = 0; n = 4), a significant correlation is present (Pearson’s r = 0.51, P = 0.041). The line of best fit is shown for Meniere’s disease in black and vestibular neuritis in grey.

4.1 AFTER trials – The effect of dizziness on locomotor aftereffects

We found that a larger locomotor aftereffect was associated with higher levels of dizziness or vertigo across both patient groups. This means that, when facing a potentially provocative situation known to have caused loss of balance (in this case the previously moving platform), the more symptomatic patients release a large “pre-emptive” postural adjustment, which is the basis of the locomotor aftereffect [48]. Association or statistical correlation however does not mean causality nor does it establish the direction of this association; does a high symptom load modulate aftereffect expression or, rather, does an exaggerated aftereffect contribute to higher levels of subjective symptoms?

Although this is a difficult question to answer, a previous study using the same locomotor paradigm in normal subjects could be of help in addressing this issue. In that study we investigated different levels of task difficulty by varying sled velocity during the MOVING trials. We observed that faster sled velocities induced higher levels of fear/anxiety and lack of confidence in being able to complete the task which, in turn, led to a larger locomotor aftereffect [18]. On this basis, and on the knowledge that anxiety and lower balance confidence levels facilitate chronic dizziness [13], our current data suggests that higher dizziness scores modulate locomotor adaptation and aftereffect expression rather than the other way round.

This is supported by our regression analysis indicating that greater trunk sway during the MOVING trials and a larger locomotor aftereffect relate to total dizziness scores and to the emotional subscores. A more detailed assessment of anxiety, fear and arousal would have been desirable and this constitutes a partial limitation of our study. Functional connectivity between limbic and postural structures in the CNS - including the parabrachial nucleus, central amygdaloid nucleus, infralimbic cortex, and hypothalamus [4, 19] - could underpin the effects that subjective symptoms and emotional aspects of balance have on adaptive locomotor control.

The ‘broken escalator’ locomotor aftereffect is unique in that despite subjects being consciously aware that the walking surface will remain stationary, they are unable to suppress the (now inappropriate) learnt motor response, mirroring the real-life experience of commuters facing a broken escalator [7, 43]. In the presence of a vestibular lesion it would seem beneficial to rely more on anticipatory, ‘open loop’ or pre-emptive postural adjustments [48]. However, our experiments show that, in doing so, patients can experience increased self-generated unsteadiness when pre-emptive postural adjustments are released inappropriately, as seen during the ‘broken escalator’ example. The finding of an increased locomotor aftereffect in subjects with high dizziness scores represents a new category of postural threat for these patients, in this case self-initiated or self-imposed.

Motor aftereffects are the result of a period of motor learning or adaptation to a perturbation. Indeed, the presence of an aftereffect is considered proof that learning or adaptation has taken place [29]. Therefore, the larger locomotor aftereffect recorded in Meniere’s disease patients could be the result of increased postural sway during MOVING trials. However, this is unlikely because vestibular neuritis patients also had significantly enhanced sway during MOVING trials and yet produced a similarly sized locomotor aftereffect to controls. In fact, even bi-labyrinthine defective patients produce a similar locomotor aftereffect to controls despite a two-fold increase in sway during MOVING trials [9]. A second possibility to explain the larger locomotor aftereffect is the influence of higher-order mechanisms and, indeed, there are several lines of evidence indicating cortical involvement in the generation of this aftereffect. Previous studies have shown that specific components of motor learning in this task are the result of open-loop predictive behaviour [42] and that the size of the locomotor aftereffect is related to the perceived risk of the task [18] and the sense of self-agency involved [44]. Moreover, the locomotor aftereffect can be modulated with transcranial cortical stimulation [26] or by observing an actor’s gait [37], likely engaging ‘mirror neuron’ systems [37]. We therefore propose that the enhanced locomotor aftereffect in very symptomatic patients reflects altered cortical processing during retention and expression of the learnt motor task, due to the influence of dizziness. Such central modulation of the locomotor aftereffect by subjective symptoms mirrors the effects of anxiety and personality traits observed upon motor learning retention, whereby participants reporting higher levels of stress and anxiety or those with extrovert personalities demonstrate greater retention in certain motor learning tasks [16, 25].

4.2 MOVING trials – Adaptation to a physically challenging task

Two findings were observed during the actual physical perturbation, that is, during the MOVING trials. Firstly, we observed that the degree of instability (trunk sway) was larger in patients with higher symptom levels. As with any other outcome in this study, instability levels were not associated with the degree of canal paresis. Thus, the motor responses to both the MOVING and the AFTER trials are less dependent on an ‘objective’ vestibular parameter such as canal paresis, indicative of how severe the vestibular damage is, than on ‘subjective’ parameters indicative of how dizzy those patients feel. This vestibulo-postural result thus parallels the lack of correlation observed between vestibulo-ocular reflex function and long-term clinical outcome [13, 38]. Clinicians should reflect on whether too much emphasis is currently being placed on pharmacological treatments aimed at modestly improving canal paresis, e.g. steroids [28], with little or no bearing on symptomatic outcome, versus other treatment resources such as rehabilitation and confidence building.

The second observation during the MOVING trials is that patients adapted to the perturbation more slowly than controls. When participants repeated the MOVING trials, the performance of healthy individuals improved rapidly by the second and third attempts as they generated more accurate predictive postural responses [8, 45] and reached a plateau [8, 42]. In patients, we observed a slower rate of adaptation compared to controls, irrespective of dizziness/vertigo symptoms or vestibular function, as also observed previously in bi-labyrinthine defective patients [9]. However, we observed a relationship between a reduced degree of adaptation to the MOVING sled and average trunk sway levels in the MOVING trials, that is, the more unsteady the patient is, the less he/she learns the task (Fig. 3). Thus, although it is not clear what dictates the degree of locomotor adaptation in this specific paradigm, multifactorial, non-specific effects seem to be at play. It seems that any abnormality in the postural control system associated with a degree of unsteadiness can interfere with the degree of adaptation to a moving surface stimulus [21, 35]. This agrees with the view that the fine-tuning required for motor adaptation is critically hindered by the presence of “noise” in the sensory and motor systems [6].

In both the BEFORE and AFTER phases, the stability of patients seems to have been maintained at the cost of gait velocity, in line with previous work [2, 27]. The slower gait velocity of patients in BEFORE and AFTER trials may thus relate to precautionary trade-off behaviour. This tallies with the enhanced locomotor aftereffect, an aftereffect that has been described as a “just in case” pre-emptive postural adjustment – anticipatory mechanisms release a small postural adjustment which, although inappropriate for a stationary support surface, would be useful ‘just in case’ the surface moved [48]. Thus, both the presence of an enhanced locomotor aftereffect and reduced gait velocity during the stationary sled trials may represent precautionary behaviour induced by dizziness or a sense of imbalance, concordant with our previous findings in bi-labyrinthine defective patients [9]. Of note, however, gait approach velocities were similar across all groups during MOVING trials, indicating that gait could be appropriately calibrated to the velocity of the moving sled if critically needed to avoid a fall. This shows that vestibular neuritis and Meniere’s disease patients are capable of adjusting gait parameters when facing a threatening external perturbation.

Another limitation of our study is that the caloric test is the sole measure of vestibular function. Additional measures of vestibular function from sinusoidal harmonic acceleration or the video Head-Impulse Test might have afforded insight into the relationship between measures of vestibular compensation and locomotor adaptation. However, it should be pointed out that the sinusoidal harmonic test [41] and Head Impulse test [31] can produce relatively normal results in Meniere’s disease and in chronic vestibular neuritis [13, 38], despite the range of locomotor responses that were observed.

In conclusion, our findings suggest that dizziness/vertigo symptoms may bias the selection of locomotor programs in favour of an over-cautious motor strategy. Symptoms interact with central motor control mechanisms and modulate output accordingly. The predictive ‘open loop’ nature of such process can occasionally lead to errors. In situations where pre-emptive or anticipatory postural adjustments are followed by a physical perturbation, such an approach might be beneficial. However, excessive postural anticipation not followed by a physical perturbation self-generates unsteadiness in patients; an unsteadiness which, as in a vicious circle, is larger in the more symptomatic patients. Such high-order mechanisms might contribute to increased unsteadiness across a variety of vestibular and neurological gait disorders.