|
|
||||||||
1Department of Otorhinolaryngology, University Hospital, Basel, Switzerland; 2Department of Biophysics, University Medical Centre St Radboud; and 3Sint Maartens Kliniek-Research, Nijmegen, The Netherlands
Submitted 20 May 2004; accepted in final form 12 July 2005
| ABSTRACT |
|---|
|
|
|---|
| INTRODUCTION |
|---|
|
|
|---|
The concept that proprioceptive input from the ankle joint alone provides the primary trigger signal for pitch-plane balance corrections has been questioned too. For example, patients with lower leg diabetic neuropathy (absent lower leg SL stretch reflexes) have adequate and practically unaltered balance correcting responses in the lower leg muscles under normal and "nulled" ankle input conditions (Bloem et al. 2000
). Therefore some researchers have postulated that the primary trigger for postural reactions originates in various sensory receptors in more proximal sites such as the knees, hips, and trunk (Allum et al. 1993
; Di Fabio 1995
; Do et al. 1988
; Forssberg and Hirschfeld 1994
; Horstmann and Dietz 1990
). Moreover, an investigation of a total leg proprioceptive loss patient caused by a dorsal root ganglionopathy (absent proprioception in both the ankle and knee joint, with severe impairment, but not total loss of proprioception at the level of the hip) provided evidence for a directionally sensitive triggering mechanism residing above the ankle and knee joints (Bloem et al. 2002
). The onsets of the balance-correcting responses in hip and trunk muscles were not delayed. This patient had significant delays in the onset of balance-correcting responses in both soleus and tibialis anterior, suggesting that knee afferent and not ankle afferent inputs may be crucial for triggering pitch plane responses (Bloem et al. 2002
). These results do not, however, provide conclusive information on the relative weighting between ankle, knee, hip, and trunk flexion proprioceptive inputs contributing to normal balance control, only that, without ankle and knee inputs, such patients do not have a sufficient redundancy of somatosensory inputs to switch to other inputs in generating a pitch-directed balance correction when perturbed. Switching between various different sensory inputs is thought to be an integral part of human balance control (Peterka and Laughlin 2004
). Furthermore, several sensory inputs, for example, proprioceptive inputs from the ankle and knee muscles and/or cutaneous inputs, would be required to distinguish between different types of support surface movements. For example, backward translation and toe-up rotation of the support surface both elicit similar profiles of ankle dorsiflexion but different knee rotations (Allum and Honegger 1998
; Allum et al. 1994; Ting and MacPherson 2004
). Thus, although one sensory input from the lower legs may play a key role in the generation of pitch-planedirected balance corrections (Ting and MacPherson 2004
), other inputs seem necessary to generate the appropriate response synergy (Allum and Honegger 1998
; Forsberg and Hirschfeld 1994
).
Because SL stretch reflexes of hip and trunk muscles are most sensitive to combined pitch and roll perturbations in specific directions (Allum et al. 2003
; Carpenter et al. 1999
) and the balance-correcting responses in these muscles have a strong roll sensitivity too, a second concept for triggering of balance corrections has evolved. This concept, with a common triggering mechanism underlying responses in sitting and standing, is organized around hip muscle responses (Forssberg and Hirschfeld 1994
). According to this concept, neural command signals use a proximal-to-distal organization of balance-correcting responses, with roll motion correction taking precedent over pitch-plane responses (Allum et al. 2003
), because trunk roll motion, after a combined pitch and roll perturbation to the body, occurs before trunk pitch motion (Carpenter et al. 1999
). In support of this concept, Carpenter et al. (1999)
revealed that ankle muscles SL stretch reflexes have a strong directional sensitivity for the pitch-plane perturbations but practically none for roll motion and raised the question of whether afferent inputs giving rise to SL ankle stretch reflexes could provide useful triggering information for responses to roll perturbations.
Inputs from the vestibular and the visual sensory systems may evoke or contribute to the triggered muscle responses with appropriate directional information. Previous research has, however, not found a major role for the visual system in either triggering or modulating balance correcting responses (Carpenter et al. 1999
; Keshner et al. 1987
). Semicircular and otolith receptors of the vestibular system would seem to provide appropriate directional triggering information by registering vertical and rotational accelerations of the head at very early onsets (within 40 ms, Allum et al. 2003
; Carpenter et al. 2001
). Nonetheless, the contribution of vestibular sensory signals to balance corrections seems to be limited to the modulation of the magnitude of postural responses. A number of our studies (Allum and Pfaltz 1985
; Allum et al. 2003
; Carpenter et al. 1999
, 2001
) have found no alterations in the onset latencies of balance-correcting responses in distal and proximal muscles in either unilateral or bilateral vestibular loss patients. An exception is the shorter latency registered in the soleus muscles of normal compared with vestibular loss subjects for toes down perturbations of the support surface, which causes the head to be accelerated vertically downward. This latency change supports a role for the otoliths in triggering postural responses during a downward fall (Carpenter et al. 2001
; Greenwood and Hopkins 1976
). The aforementioned findings add weight to the hypothesis that both pitch and rolldirected balance-correcting responses may well receive key triggering information from sites more proximal than the ankle joints and more caudal than the vestibular system. While some evidence described above indicates a major roll of afferents underlying hip and trunk SL stretch reflexes in triggering roll-directed balance corrections in other nonstretched muscles, differentiating between knee and hip proprioceptive afferents as the key proprioceptive triggering source for pitch-directed balance corrections remains difficult.
It has been established that when a multidirectional perturbation to stance is corrected, a multi-segmental strategy is used that involves movements about the knees, hips, and lower vertebral column in addition to ankle joint motion (Allum et al. 2002
, 2003
; Carpenter et al. 1999
; Grüneberg et al. 2004
). Because the response dynamics of the legs and trunk have a different timing in the pitch and roll planes (Allum et al. 2002
, 2003
; Carpenter et al. 1999
), we assume that neural command signals must employ a temporal separation between the roll and pitchdirected synergies to correct for combined roll and pitch to stance perturbations. Adequate balance corrections in these two planes must be processed and perhaps generated sequentially as more rapid trunk roll than pitch movements are required to compensate for falls with both roll and pitch components (Carpenter et al. 1999
). Supporting evidence for this notion comes from recent studies that showed changes in the temporal coupling between trunk roll and pitch motion with sensory deficits (Bloem et al. 2002
; Carpenter et al. 2001
), ageing (Allum et al. 2002
) and artificially imposed trunk stiffness (Grüneberg et al. 2004
). In summary, these findings add weight to previous suggestions (Henry et al. 1998
; Winter et al. 1996
) that the execution of balance-correcting responses in the pitch and roll planes must be organized separately. To date, there have been no studies that have examined postural reactions to multi-directional perturbations by separately applying roll (or lateral) and pitch (or anterior-posterior) components of the perturbation to tease out details of the organization of neural commands in this regard. We chose to study the separation of the neural control of roll and pitch body motion by employing first a roll stimulus delay with respect to pitch equal to the delay of stretch reflexes in leg muscle (
50 ms), and second, a roll delay of 150 ms based on differences in the timing of peaks in trunk roll and pitch angular velocities when no delay is present. We were also interested in determining which channels of sensory information might provide triggering and directional information useful for generating appropriate balance corrections. The hypothesis we examined here was that pitch motion would be controlled by motion mainly about the ankle and knee joints and roll motion by motion about the hip and lumbro-sacral joints with little interaction between the two types of motion control at ankles and trunk. We assumed that sensory information present in muscle afferents would be observable in SL reflexes of muscles stretched or released by the perturbations we applied.
| METHODS |
|---|
|
|
|---|
Outcome measures
Biomechanical and EMG outcome measures were collected using previously described techniques (Allum et al. 2002
; Carpenter et al. 1999
). To record EMG signals, pairs of silver-silver chloride electrodes were placed
3 cm apart along the muscle bellies of left tibialis anterior, left soleus, left peroneous longus, left and right gluteus medius, left paraspinals at the L1L2 level of the spine, and left abdominal muscle, oblique externus. EMG preamplifier band-pass filtering was over 0.7 Hz to 2.5 KHz. Pairs of electrodes and lead lengths assigned to individual muscles were not changed between subjects throughout the experiments. Overall gains (including subsequent amplification and filtering) were also kept constant at 1020,000. The latter value was channel dependent.
Support surface reaction forces of both feet were measured from strain gauges embedded within the rotating support surface. The strain gauges were located under the corners of the plate supporting the feet. From forces recorded perpendicular to the platform by the strain gauges of the left foot and the distances to the center of ankle joint rotation, the anterior-posterior (AP) and mediolateral (ML) ankle torques were calculated for the left foot. Because a difference in strain gauge measures was used for torque calculations, an influence of the platform mass on the torque measurements was negligible. A similar system measured forces and torques applied by the right foot. Angular velocity of the upper trunk in the pitch and roll planes was collected using Watson Industries transducers (±300°/s range) mounted onto a metal plate that hung at the level of the sternum from shoulder straps that wrapped around the shoulders back and chest. To measure lower leg angle in the roll and pitch planes, a lightweight metal rod was fixed with an adjustable strap to the lateral aspect of the left tibia,
4 cm below the level of the lateral condyle. The rod was connected to a potentiometer located on the pitch axis of the platform. Estimates of ankle dorsiflexion and inversion were calculated based on the differences between these angles and the support surface pitch and roll rotation, respectively. The latter two angular variables were measured with potentiometers mounted on the axes of support surface rotation. Head linear and angular accelerations (see also Allum et al. 2003
) were computed from the outputs of four dual axis linear accelometers (Entran), with ranges of ±5g, mounted at 90° separation on a lightweight, adjustable headband. The only variable measured for this study was the roll angular acceleration (difference of the vertical linear accelerations at the level of the ears).
Procedure
The subject's feet were lightly strapped into heel guides fixed to the top surface of the dual-axis platform that rotated in the pitch (forward-backward) and roll (lateral tilt) directions. The guides were adjusted in the AP direction to ensure that the ankle joint axis was aligned with the pitch axis of the rotating platform. The height of the pitch rotation axis above the support surface was equal to average height of the ankle joint above the soles of the feet. The roll axis had the same height as the pitch axis above the support surface and passed between the feet. Just before the experiment, subjects were asked to assume their "preferred" standing posture with the arms hanging comfortably at their sides. At each individual's preferred-stance position, we measured the low-pass filtered (5 Hz) sum of the AP torques measured from each foot. This measurement was treated as the reference value for preferred-stance for the remainder of the experiment.
Subjects were presented 99 perturbations in three series of 33 perturbations each with a rest of 5 min between each series. The order of the series presentation was counterbalanced among the subjects. The first trial of each series was excluded from further analysis to reduce habituation effects entering the data (Keshner et al. 1987
). The remaining 96 perturbations consisted of randomized order of eight different perturbation directions and three types of delays of the roll perturbation with respect to pitch (0, 50, and 150 ms). All perturbations had one velocity (60°/s) and a constant amplitude of 7.5°. Previous experiments (Allum et al. 1993
, 2003
) determined that these perturbation parameters were sufficient to elicit a response synergy that included as prerequisite an active trunk movement for pitch perturbations as determined by the presence of trunk flexor muscle activity (see RESULTS).
Defining 0° as a forward pitch rotation and 90° as a right roll rotation, the eight perturbation directions were 23, 68, 113, 158, 203, 248, 293, and 338°, with a direction separation of 45°. Each delay-direction combination was presented four times to each subject.
Each perturbation was preceded by a random 5- to 20-s delay. During this period, subjects were asked to monitor an oscilloscope, which was located at eye level,
1 m in front of them. This oscilloscope displayed on-line the low-passfiltered AP torque, which was measured as described above. Using this visual feedback, subjects were required to maintain AP ankle torque within a range of ±5 Nm from their preferred-stance reference value. The 5- to 20-s interstimulus delay was initiated automatically once the platform had returned to its original prestimulus position and the subject had regained and maintained his preferred vertical position as monitored by AP ankle torque reading. In response to each rotational perturbation, subjects were instructed to recover their balance as quickly as possible using in-place reactions. Because of the foot straps, stepping reactions were not possible. Three handrails (generally 80 cm high but adjustable to hand height of each subject) were located at a distance of 40 cm to the sides and to the front of the platform center. Subjects were informed they were allowed to grasp the handrails, if needed, but they did not need to. A spotter was always present to lend support in case of a fall, but no falls, or near falls, occurred.
Data analysis
All EMG and biomechanical recordings were initiated 100 ms before perturbation onset and had a sampling duration of 1 s. EMG recordings were band-pass analog filtered-between 60 and 600 Hz, full wave rectified, and low-pass filtered at 100 Hz before sampling at 1 KHz (a verification of this filtering for tracking EMG response onsets and area calculations is found in Gottlieb and Agarwal 1970
, 1979
). All biomechanical data were sampled at 500 Hz after passing through anti-aliasing filters and digitally low-pass filtered off-line at 25 Hz using a zero phase-shift 10th-order Butterworth filter.
After analog to digital conversion of the data, all biomechanical and EMG signals were averaged off-line across each perturbation direction/delay combination. Zero latency was defined as the first deflection of ankle rotation velocity trace above a level of 5% of the maximum stimulus velocity and did not vary with direction or subject. Subject averages were pooled to produce population averages (of 76 single traces) for a single direction and delay (as shown in Figs. 1, 2, and 7). Peak trunk roll and pitch angular velocity was calculated as the absolute maximum velocity over the intervals between 0 and 500 ms. To create the angular displacements plots of Fig. 2, angular velocity traces of the trunk were integrated off-line using trapezoid integration. Peak head roll angular acceleration was calculated as the maximum of the first peak. Ankle torque changes were calculated between 160260 and 280380 ms.
|
|
|
With this technique, we were able to capture variations in balance-correcting responses across subjects, muscles, and roll stimulus delays using the same integration areas for all subjects.
Once we had determined the onsets of balance-correcting responses, we examined EMG activity before those onsets to determine the onset of SL stretch reflex activity using a similar 10% threshold technique but based on the first derivative of the EMG signal.
Our primary statistical analyses concerned between-condition comparisons for delay effects. To examine the effects of different perturbation directions and delay conditions, we used a two-way ANOVA model for repeated measurements within subjects after testing that the EMG and the biomechanical data were normally distributed with a Kolmogornov-Smirnov test. Significant main effects, delay condition, and interaction effects with direction were further explored using post hoc comparisons with a Bonferroni correction to account for the effect of comparing three delay conditions at once. Differences in the EMG onset, maximum response time (tmax), and duration of the response (Intm 0, 50, 150) with stimulus delay were tested with one-wayANOVAs and t-test with Bonferroni corrections. Results with P < 0.05 were considered significant.
| RESULTS |
|---|
|
|
|---|
Biomechanical responses
Figure 2 shows how the delayed roll movements of the support surface perturbations altered the roll responses of the lower leg, trunk, arm, and head in a consistent manner. Figure 2 shows population traces for several biomechanical variables measured in the pitch and roll planes in response to a slightly backward and right rotation of the support surface in the direction of 113° (90° is a pure right rotation). Both delay conditions induced a major difference in the timing of the biomechanical responses of the roll movements at the ankle joint, trunk, and head. The peak amplitude of these roll responses at each perturbation direction was not changed with stimulus delay. Up to the time-point of peak ankle roll angle, the pitch and roll ankle responses followed the trajectory of the support-surface rotation (Fig. 2).
The effects of the 50- and 150-ms delay conditions are shown in Fig. 3 for pitch (dorsi-flexion) and roll (inversion) ankle angular velocity. The right polar plot in Fig. 3 shows that there was no significant effect of the delay on the amplitude of peak ankle roll velocity [F(2,36) = 1.3, P = 0.286]. There was, however, a consistent, but small, change to the amplitude of peak ankle pitch angular velocity [F(2,36) = 6.5, P = 0.004]. The significant changes occurred mostly for roll-oriented perturbations such as for the 68 and 248° directions (P < 0.05; see left polar plot in Fig. 3). For example, the average values of peak pitch velocity for the 248° perturbation direction were 38.5, 32.5, and 31.2°/s for 0-, 50-, and 150-ms delays, respectively, with SD of 6.1, 1.7, and 1.4. From these values, it is clear that the mean for 0 ms is larger. The 50- and 150-ms delay of the platform roll relative to the platform pitch induced a major shift in the time to peak ankle roll angular velocity [F(2,36) = 19.1, P < 0.001]. Across rightward directions of 68 and 113°, for example, the peak occurred on average at 79 ms for the 0-ms delay, 126 ms for the 50-ms delay, and 226 ms for the 150-ms delay (Fig. 3, bottom right). Thus the stimulus delays of 50 and 150 ms resulted in a consistent 47- and 147-ms delay to peak ankle roll velocity. The time to peak ankle dorsi-flexion (pitch) velocity changed for some directions with the stimulus delays (Fig. 3, bottom left); nonetheless, the changes were always <25 ms, being slightly larger for forward and roll directions (23 and 68°) and slightly less for backward and roll directions (113 and 158°) in the left (uphill) leg (Fig. 3) and therefore not consistently increased with stimulus delay. The resultant effect of stimulus delays on the vector direction of ankle angular velocity is shown by the vector polar plots in Fig. 3. These plots show the vector directions of angular velocity measured at 79 and 226 ms. At 79 ms, when ankle roll angular velocity peaks on average for the 0-ms delay, the vector directions of ankle velocity for the 0-ms delay are directed opposite the perturbation direction, but for the 150-ms delay, these vectors at 79 ms were shifted toward a purely pitch orientation (left vector polar plot in Fig. 3). At 226 ms, when the ankle roll angular velocity peaks for the 150-ms delay, the vector directions for ankle angular velocity with the 150-ms delay are clearly roll oriented (right vector polar plot in Fig. 3). The vector directions of ankle velocity for the 0-ms delay at 226 ms were influenced by the beginning of the roll-directed balance correction. For example, as shown in Fig. 2, the left ankle dorsi-flexed again (and presumably the knee, because it was the lower leg rotation that caused the ankle dorsi-flexion) at 200 ms for the 0-ms delay and later for the 50- and 150-ms delays. Using a 3°/s threshold for the start of this second dorsi-flexion, this started at 197 ± 30, 198 ± 35, and 267 ± 37 (SD) ms (for 0-, 50-, and 150-ms delays, respectively) for the plots shown in Fig. 2 (113° perturbation direction). Only the onset of 264 ms for the 150-ms delay had a significant (P < 0.001) delay of 64 ms with respect to the onsets at 0 and 50 ms.
|
|
3°; Fig. 2) also shifted with delay. Using a return to <3°/s as a criterion, these times were 266 ± 54, 315 ± 68, and 382 ± 54 ms for the 0-, 50-, and 150-ms delay conditions for the 113 directions (Fig. 2). The differences in these mean times (49 and 116 ms) were significant (P < 0.05). There were some changes in trunk peak pitch velocity and time to peak pitch velocity. The profile of trunk pitch velocity was somewhat greater for the slightly backward left and right roll directions for the zero delay condition and otherwise unaltered (see Figs. 2 and 5). Changes only occurred for perturbations with a small pitch but large roll component (113°, as in Fig. 2, and 248° directions). Thus the pitch peak velocity was reduced [F(2,36) = 19.4, P < 0.001], specifically for the 150-ms delay condition for 113 and 248° (P < 0.05), and the time to peak pitch velocity of the trunk was altered [F(2,36) = 4.4, P = 0.019] being delayed for the 68° direction and earlier for the 113° direction (P < 0.05) with respect to the time for 0-ms delay (Fig. 5).
|
The same trend described above for the stimulus-induced roll movements of the ankles and trunk was also found for roll accelerations of the head (Fig. 6, top plot). As shown in Fig. 6, there was no significant difference in the peak amplitudes of head roll acceleration for the different delay conditions [F(2,36) = 0.9, P = 0.412]. The time to peak head roll acceleration was delayed for both delay conditions [F(2,36) = 66.7, P < 0.001] and, for all directions, peak times showed significant differences with respect to each delay condition (P < 0.001). The mean time to peak head acceleration was 108 ± 25.2 ms for rightward perturbations for 0 delay, 161 ± 30.1 ms for the 50-ms delay (a difference of 53 ms), and 250 ± 31.7 ms for the 150-ms delay (a difference of 142 ms with respect to the 0-ms peak times).
|
EMG activity
The delayed roll stimulus motion caused several effects on the EMG responses as shown in Fig. 7. This figure shows population EMG traces for a slightly backward and right rotation (113°) of the support surface (the same direction used for Fig. 2). These EMG traces show our general finding that the delay in trunk roll motion caused a corresponding delay in the onset of the most prominent balance-correcting muscle activity in trunk muscles (Fig. 7, Intm 150) consistent with shifts in the timing of trunk stabilization poststimulus. A smaller, separate, and earlier part of the balance-correcting response in trunk muscles most easily observed with the 150-ms delay was not delayed (Fig. 7, Intm 0). In contrast, the delayed ankle and trunk roll motion caused only a trend for small shifts in balance-correcting activity in lower leg muscles (Figs. 79), consistent with the small shifts in the timing of the leg-flexion response to roll.
|
45 ms on the soleus and peroneus longus traces in Fig. 7 mark the approximate onset of SL stretch reflexes in these muscles. It can be observed in Fig. 7 that these SL reflex responses are not altered by delays in stimulus roll. In contrast, the SL unloading reflexes in gluteus medius and small SL stretch reflex in obliques externus (onset marked by vertical arrows at 30 ms in Fig. 7) were clearly delayed by the 150-ms delayed roll stimulus (compare upward and downward pointing vertical arrows on EMG traces of these muscles in Fig. 7). We have previously shown that movement of the upper leg relative to the pelvis is consistent with SL stretch reflexes in gluteus medius, and movement of the trunk relative to the pelvis is consistent with SL stretch reflexes in paraspinals (Allum et al. 2003
|
When EMG activity was measured over an SL stretch reflex response activity period (3070 ms) in trunk and hip muscles, clear responses were seen for the 0-ms delay in all trunk and hip muscles. However, EMG activity over this interval was practically abolished for the 50-ms delay stimulus and absent for the 150-ms delay. The left gluteus medius, which is stretched by right tilt perturbations, and the right gluteus medius, which is unloaded as the pelvis rotates more right than the legs (Allum et al. 2003
; Fig. 8), showed practically no activity change with respect to prestimulus activity for any stimulus with roll delays 80 ms after stimulus onset. Hence a significant decrease in SL reflex activity 3070 ms after stimulus onset occurred with roll delay [F(2,36) = 6.6, P = 0.003]. Post hoc comparisons for all directions revealed a significant change over the period 3070 ms with delay (P < 0.001). Left paraspinals, which are stretched by leftward perturbations as a result of trunk roll right and pelvis roll left, showed some reflex activity before 80 ms for the 50-ms delay stimulus (Fig. 8), but this was considerably reduced with respect to the 0-ms delay stimulus [F(2,36) = 5.5, P = 0.003]. Post hoc comparisons for all left perturbations showed significantly reduced (P < 0.01) activity in left paraspinals between 30 and 70 ms with each delay. Like paraspinals, the left external abdominal oblique was stretched by a support surface roll left. Activity over the 30- to 70-ms interval in the external abdominal oblique was negligible for delayed roll stimuli [F(2,36) = 41.4, P < 0.001]. Post hoc comparisons revealed significant decreases for the 113 and 203° (P < 0.05) and 248, 293, and 338° (P < 0.01) directions.
Lower leg balance-correcting responses mostly retained their timing and amplitude characteristics despite roll delays in leg and trunk movements. The top of Fig. 9 shows the amplitude of the area of the balance-correcting responses measured over an interval based on the time of peak activity of the responses for all three delay conditions (Intm0, Intm50, and Intm150). As these polar plots show, no significant differences were found in the amplitudes with delay in any direction for the peroneus longus or soleus muscles [soleus: F(2,36) = 0.04, P = 0.959; peroneus longus: F(2,36) = 1.5 P = 0.234]. For the tibialis anterior, a small overall effect emerged [F(2.36) = 3.9; P = 0.027]. However, for the backward directions (for which the balance corrections and the effects of the delays were largest), we found no significant effects in post hoc analysis (P > 0.05). The arrows on the polar plots in the top of Fig. 9 indicate the direction of maximum activity with respect to the support surface perturbation direction. The changes in this direction inducing the maximum response were minimal with delay (2° in tibialis anterior, 6° in soleus, and 2° in peroneus longus) between the 0- and 150-ms delay stimuli. We also used the same integration interval used for the 0 delay responses to determine the areas of balance-correcting responses falling within this area for the 50- and 150-ms delay stimuli (Fig. 9, bottom row). In other words, a subject specific fixed interval, that for the 0-ms stimulus, was employed across all delays. Over this interval (Intm 0 in our nomenclature), the responses of the lower leg were also unaltered in amplitude for the soleus [F(2,36) = 1.0, P = 0.372], slightly altered in amplitude for the peroneus longus [F(2,36) = 15.9, P < 0.001], for which the direction 293° showed a significant decrease (P < 0.001), and also slightly decreased in amplitude for the tibialis anterior [F(2,36 = 4.3, P = 0.02] for the same direction (P < 0.05). Note, however, that this direction elicited a very small response (Fig. 9). The directions of maximum responsiveness also remained basically unaltered when this integration interval was used. Thus in general, only small changes in ankle muscles balance correcting response amplitudes with delay were seen for perturbation directions causing considerable balance-correcting activity.
Although minor changes in the response onsets, time of the peak response, and the response interval of the balance-correcting responses in the lower leg were noted with a tendency for these times to increase with roll stimulus delay (Fig. 10), this trend was not significant. Exceptions to this general finding were the onset of the soleus [F(2,54) = 3.9, P = 0.025], which was less for the 0-ms delay with respect to 150 ms (P < 0.05), the time to peak of the response (tmax) of the peroneus longus [F(2,54) = 4.1, P = 0.021], for which the time for the 50-ms delay was marginally less than for the 150-ms delay (P < 0.05), and the length of the balance-correcting response interval for the tibialis anterior [F(2,54) = 5.2, P = 0.009], which was significantly different between the 0- and 150-ms delay (P < 0.01). However, as reported above, these differences did not lead to any effects on the amplitudes of response areas of the balance correcting responses (Fig. 9).
|
50 and
120 ms. Similar delay differences were observed for the onsets of the main balance correcting response between 0-, 50-, and 150-ms stimulus delays (Fig. 10). The shift of the onset of the hip and trunk balance-correcting responses caused by the delay in trunk roll motion was reflected in the amplitudes of the responses and had an effect on the directional sensitivity. Figure 11 shows the changes in amplitude and directional sensitivity of these muscles with stimulus direction and delay. The magnitude of the response reduction with delay depended on the way EMG area over the balance-correcting interval was measured. When the response EMG areas were calculated based on an interval around the time of peak response for each condition (Intm0, Intm50, and Intm150), the amplitude was reduced significantly across delay conditions (Fig. 11, top row) in the paraspinals [F(2,36) = 6.1, P = 0.005], specifically for the directions of 158 and 203° (P < 0.05), and in the external abdominal oblique [F(2,36) = 21.2, P < 0.001] for the direction of 293° (P < 0.05). The trend for a reduction in the gluteus medius was not significant. Correspondingly, the direction of maximum response sensitivity changed little in the gluteus medius (252° with no delay, 246° with 150-ms delay) but became more roll-oriented in the paraspinals (changing from 136 to 108°; Fig. 11). All of these muscles, however, showed even more significant amplitude reductions when areas were recalculated based only on the interval centered around the peak response for the 0-ms delay stimulus (Intm0). These changes are shown in the middle row of Fig. 11. Using this measurement interval, the balance-correcting responses in the gluteus medius were reduced [F(2,36) = 34.3, P < 0.001], which is very apparent for all leftward directions (P < 0.0001) in Fig. 11, as were those of the paraspinals [F(2,36) = 20.7, P < 0.001], especially for more right roll directions of 68 and 113° (P < 0.01). Figure 7 shows how the EMG activity in trunk muscles became split into a weaker early response and a stronger more prominent delayed response with the 150-ms stimulus delay. The direction of the maximum response activity of this first, weaker response before main balance correcting with an onset >200 ms became very pitch-oriented (Fig. 11, middle). This direction was 189° in the gluteus medius and 175° in the paraspinals. Responses in the external abdominal oblique [F(2,36) = 18.6, P < 0.001] were also reduced when the Intm 0 interval was used across all delays. The post hoc tests revealed that there was a significant difference in response amplitudes for this muscle for the perturbation directions of 248 and 293° (P < 0.005).
|
| DISCUSSION |
|---|
|
|
|---|
Our previous finding with multi-direction perturbations of stance were first that, when the body was tilted with a combined pitch and roll perturbation, the trunk roll was earlier than trunk pitch and followed by flexion of the uphill leg. Second, when the body was tilted in the pitch plane, the trunk pitch still occurred later than when the trunk rolled for a pure roll perturbation. However, considerable early knee motion was observed before trunk motion when the support surface was tilted toe-down and early ankle motion when the tilt was toe-up (Allum et al. 2003
; Carpenter et al. 1999
). These findings led us to develop the hypothesis in humans that pitch motion would be controlled by motion mainly about the ankle and knee joints and roll motion by motion about the lumbro-sacral joints, with little interaction between the two even for a combined pitch and roll perturbation (Allum et al. 2003
). This study was an attempt to investigate this hypothesis concerning neural command signals to multi-directional perturbations in humans by applying the roll and pitch components of the support surface time lagged with respect to one another. We were careful to apply a pitch stimulus of 60°/s, approximately twice as fast as that necessary to elicit balance correcting responses in trunk muscles (Allum et al. 1993
), to be sure that, according to the criteria of Runge et al. (1999)
, a hip strategy was part of the response to pitch.
By shifting the roll stimulus with respect to the pitch component, our results showed that the responses in the hip and trunk muscles could be split into two components: one predominant response that shifted with the roll delay and a smaller response that did not shift. In the ankle muscles, a trend for a small shift in the main balance-correcting response occurred, but it neither reached statistical significance nor gave the appearance of two components. While these results are not conclusive and strictly can only be applied to our experimental conditions, we interprete these results as support for the concept of pitch and roll balance corrections being handled separately by neural command centers. Our results also support the idea of AP motion control predominantly by the ankle muscles and ML control by the hip and trunk muscles; however, our results add important caveats. The presence of the undelayed hip and trunk muscle activity indicates that this activity is part of the ankle and hip/trunk synergy controlling pitch. Furthermore, a trend for a delay in ankle flexor muscle activity (tibialis anterior and peroneus longus) indicates that a leg flexor response is part of the synergy controlling roll.
Similar to our use of nulled ankle input perturbations during pitch-plane perturbations (Bloem et al. 2000
, 2002
), our different types of roll delays with respect to the pitch perturbation were also designed to differentiate between different triggering mechanisms for balance corrections, specifically those at the ankle joint and at more proximal joints, as well as the characteristics of neural command centers controlling the body's roll and pitch motion. With nulled ankle inputs, we were able to suppress SL stretch reflexes in ankle muscles, yet show that balance corrections were present in the very same muscles with minor reductions in amplitude (Bloem et al. 2000
). Therefore we concluded that ankle inputs could not be the crucial triggering source for these responses, but possibly knee and/or hip joint inputs could (Bloem et al. 2000
). Our rationale with the use of 50- and 150-ms delays of the roll perturbation with respect to that of pitch was similar. If we could show that hip and trunk reflexes were delayed corresponding to the delay in roll, yet leg muscle balance corrections not altered in a similar fashion, this would strengthen our arguments with regard to independent pitch control of body motion triggered by knee inputs. In addition, if balance corrections in trunk muscles were delayed with the roll stimulus, regardless of the earlier presence of ankle and knee inputs with onset of the pitch stimulus, this evidence would support our arguments concerning a crucial triggering role for balance correcting responses using muscle and joint inputs from the hip and lumbro-sacral joints. The results of these experiments support these hypotheses, despite a number of caveats that we will discuss. Our primary conclusion is that the neural commands must employ spatio-temporally separated roll and pitch balance correcting commands because the response dynamics of legs and trunk are very different in the roll and pitch planes, as we have documented here and previously (Allum et al. 2003
; Carpenter et al. 1999
).
Biomechanical separation of pitch and roll motion of the body
The backbone to our argument about the need for separate roll and pitch commands is based on how the body moves when it is perturbed. Our previous work has shown that when the support surface tilts into roll, the legs first act like two pistons to drive the pelvis in the same direction as the tilt and the trunk in the opposite direction (Allum et al. 2003
). This motion occurs with very little early ankle and knee flexion. In contrast, pitch displacements cause very early rotation of the ankle and knee joints, with the knees becoming locked when the support surface tips backward, thereby extending the knee joints (Allum et al. 2003
). These two different types of planar motion result in differences in the timing in trunk roll and pitch motion. As Figs. 2 and 5 show, the trunk roll velocity peaks around 140 ms normally (for the 0-ms delay) and pitch motion around 100 ms later. The amount of this roll motion changes with age. First, the amount becomes significantly less than that of 20 yr olds around the age of 45. After the age of 60, the initial roll is in the same direction as the perturbation (Allum et al. 2002
). In pathologically stiff persons, the trunk roll is even greater in the same direction as the tilt perturbation (Bloem et al. 2002
) than that of the elderly. These changes in trunk roll characteristics with age are not accompanied by changes in the amount and velocity of pitch motion of the trunk. To cope adequately with the basic timing differences in trunk pitch and roll motion and with the changes in this timing with aging, separate pitch and roll commands would seem essential.
These results suggest that trunk roll and pitch motion control commands can be separated. By delaying the roll motion of the support surface with respect to pitch, we were able to shift the time-point of trunk roll motion (and that of the head) by an amount equal to the roll delay. Similar changes in pitch motion of the trunk were not observed. Some changes in trunk pitch were inevitable (see Figs. 2 and 5) because a roll perturbation induces some trunk pitch particularly for roll perturbation with a backward tilt component (Carpenter et al. 1999
). Prior research has suggested this occurs with the uphill knee flexion response (Allum et al. 2003
). The changes in trunk pitch that did occur were small and different for different perturbation directions and not as profound as the changes in roll motion with successive delays in the roll stimulus. These small changes were consistent with the leg flexion response as observed in both the small changes in onset of ankle flexion around 200 ms and in tibialis anterior and peroneous longus activity with roll delay. The overall effect was that, at
140 ms, when normal motion of the trunk is solely roll oriented, this motion became only pitch oriented when the roll stimulus was delayed 150 ms. Whereas at
240 ms, when trunk motion is normally pitch oriented, motion became opposite the direction of perturbation for the 150-ms delay stimulus.
Triggering of balance corrections
Before discussing the characteristics of pitch and roll synergies, it is instructive to consider how previous research and these experiments shed light on the triggering signals for these synergies.
Lower leg somatosensory information seems to be important, but not essential, for triggering balance corrections. Neuropathy patients with absent SL stretch reflexes in lower leg muscles (presumably indicating that Ia muscle afferents are not functioning normally) show small (2030 ms) delays in the onsets of balance-correcting responses in leg muscles (Bloem et al. 2000
; Inglis et al. 1994
). In addition, those with cutaneous deficits in the feet showed greater individual variability in these onsets (Simmons and Richardson 2001
), possibly suggesting greater difficulty in switching to alternative sources of triggering information when cutaneous and muscle afferents are deficient. This difficulty may arise as well because cutaneous afferents information from the feet has been noted to be ideal for detecting the direction of support surface perturbations (Ting and MacPherson 2004
).
On the other hand, it was proposed that, in contrast to previous claims (Fitzpatrick et al. 1992
), lower leg inputs have a relatively minor role in triggering balance corrections (Bloem et al. 2000
). When deprived of these inputs either as a result of disease (polyneuropathy) or experimentally balance corrections in leg muscles were minimally changed (Bloem et al. 2000
). It was proposed instead that upper leg (knee and hip flexor muscle receptors and joint afferents) rather than those of the lower leg might provide proprioceptive trigger signals for pitch-directed balance corrections (Allum and Honegger 1998
). A study on a total leg proprioceptive loss patient (Bloem et al. 2002
), with absent proprioception at the level of the ankle and knee joint (but questionable loss at the hip), provided supporting evidence for this upper leg triggering proposal. This study supports this concept because, when early SL stretch reflexes were significantly reduced in hip and trunk muscles with delayed trunk roll, onsets of leg muscle balance corrections showed only a nonsignificant trend to be delayed. Previously it was proposed that afferents underlying the predominantly roll-sensitive SL stretch reflexes in laterally acting hip and trunk muscles could provide, albeit