Direction-specific postural instability in subjects with Parkinson's disease

Experimental Neurology 193 (2005) 504 – 521 www.elsevier.com/locate/yexnr

Direction-specific postural instability in subjects with Parkinson’s disease Fay B. Horaka,b,c,*, Diana Dimitrovaa, John G. Nutt b,c a

Neurological Sciences Institute, Oregon Health and Science University, Portland, OR 97006-3499, USA b Department of Neurology, Oregon Health and Science University, Portland, OR 97006-3499, USA c Department of Physiology and Pharmacology, Oregon Health and Science University, Portland, OR 97006-3499, USA Received 20 April 2004; revised 28 September 2004; accepted 8 December 2004 Available online 19 February 2005

Abstract The purpose of this study was to determine whether and why subjects with Parkinson’s disease (PD) have greater instability in response to specific directions of perturbations than do age-matched control subjects and how instability is affected by stance width. This study compared postural responses to 8 directions of surface translations in PD subjects and age-matched control subjects while standing in a narrow and wide stance. PD subjects were tested in their practical OFF state. A postural stability margin was quantified as the difference between peak center of pressure (CoP) and peak center of mass (CoM) displacement in response to surface translations. The control subjects maintained a consistent stability margin across directions of translations and for both narrow and wide stance by modifying rate of rise of CoP responses. PD subjects had smaller than normal postural stability margins in all directions, but, especially for backwards sway in both stance widths and for lateral sway in narrow stance width. The reduced stability margin in PD subjects was due to a slower rise and smaller peak of CoP in the PD subjects than in control subjects. Lateral postural stability was compromised in PD subjects by lack of trunk flexibility and backwards postural stability was compromised by lack of knee flexion, resulting in excessive displacements of the body CoM. Stability margins in PD subjects were related to their response on the pull test in the Unified Parkinson’s Disease Rating Scale. Thus, PD patients have directionally specific postural instability due to an ineffective stiffening response and inability to modify their postural responses for changing postural demands related to direction of perturbation and initial stance posture. These results suggest that the basal ganglia, in addition to regulating muscle tone and energizing postural muscle activation, also are critical for adapting postural response patterns for specific biomechanical conditions. D 2004 Elsevier Inc. All rights reserved. Keywords: Parkinson’s disease; Posture; Balance; Equilibrium

Introduction Although a backward pull at the shoulders is used to identify postural instability in patients with Parkinson’s Disease (PD), it is not clear that they are more unstable in the backward direction than in forward or lateral directions (Allum et al., 2002; Greenspan et al., 1998). Assessing dynamic postural stability across different directions of perturbations could help determine whether and why there * Corresponding author. Neurological Sciences Institute, Oregon Health and Science University, OHSU West Campus- Building 1, 505 NW 185th Avenue, Beaverton, OR 97006-3499, USA. Fax: +1 503 418 2501. 0014-4886/$ - see front matter D 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.expneurol.2004.12.008

may be a directional preponderance of falls in PD (Grimbergen et al., 2004). Based on biomechanical principles, it is generally assumed that standing humans are most unstable in response to backward body displacements since it is more difficult to exert dorsiflexion, than plantarflexion, torque about the ankles (Winter et al., 1996). However, our previous studies of multidirectional surface translations in young healthy subjects showed that the nervous system may compensate for biomechanical constraints by increasing the magnitude of muscle activation and surface reactive forces triggered in response to backward body displacements (Henry et al., 1998b, 2001). These previous studies also showed that healthy young subjects immediately increase the magnitude

F.B. Horak et al. / Experimental Neurology 193 (2005) 504–521

of their proximal postural responses to lateral displacement when the postural demands are increased by standing in narrow stance. These, and other studies, suggest that the healthy nervous system takes into account initial biomechanical conditions to flexibly modify postural responses to maintain stability (Macpherson et al., 1988; Nashner and Cordo, 1981). Unlike healthy subjects, PD subjects have difficulty modifying the magnitude and patterns of postural responses for changes in postural demand. The importance of the basal ganglia for set-dependent adaptation of postural movement patterns for changes in conditions has been shown previously for changes in arm support, changes from standing to sitting posture, surface rotation versus translation, and instructions to PD and control subjects (Chong et al., 1999, 2000; Diener et al., 1987; Horak et al., 1992; Schiepatti and Nardone, 1991). Recently, we found that PD subjects also do not modify the magnitude of their muscle activity or resulting direction and magnitude of horizontal reactive forces when changing stance width (Dimitrova et al., 2004a,b). The effects of this postural inflexibility on the biomechanics of postural stability in different conditions are unknown. The purpose of this study was to determine whether and why PD subjects have greater instability in response to perturbations in specific directions than do age-matched controls and how this direction-specific instability is affected by stance width. We compared how PD subjects and age-matched control subjects adapt their kinetic and kinematic postural responses to different directions of surface perturbations and to different stance widths. A better understanding of postural deficits in PD subjects may provide insight into the role of the basal ganglia in postural control. For example, if the basal ganglia are particularly important for axial control of trunk (axial) coordination, we would expect PD subjects to have more deficits in control of lateral, postural stability because lateral stability results primarily from hip and trunk control, whereas AP stability results primarily from ankle control (Henry et al., 1998b, 2001; Winter et al., 1996). Furthermore, we would also expect larger postural deficits in narrow stance because responses to lateral surface translations involve primarily weighting and unweighting of the legs with very little trunk movement when the legs are far apart, whereas postural responses require large lateral flexion of the trunk when the legs are close together (Henry et al., 2001; Winter et al., 1996). Biomechanical constraints related to the shape, size, and strength of the foot, limitations of joint motion, and stiffness and flexibility of body parts necessarily interact with neural constraints imposed by basal ganglia pathology to affect postural stability differently across perturbation directions. To restore a falling body to stable equilibrium, the centerof-foot pressure (CoP) must move in front of the falling center-of-mass (CoM) to return the CoM safely within the base of foot support (Winter et al., 1990; Yang et al., 1990). The CoP is the location of the net reactive forces at the

505

surface (Horak and Macpherson, 1996). In standing humans, the difference between the peak CoP and peak CoM has been used as a quantitative measure of the bfunctional stability marginQ (Winter et al., 1996), which is a functional measure of dynamic stability. The closer that the surface projection of the CoM approaches the CoP, the more likely is an individual to loose equilibrium and to fall or stumble (Yang et al., 1990). In response to forward body sway induced by backward surface translations, our previous study showed that the difference between the peak CoP and the peak CoM in PD subjects was significantly smaller than that in age-matched control subjects (Horak et al., 1996a). This reduced stability in PD subjects occurred despite slower velocity of CoM displacements due to increased passive stiffness. These results suggest that postural bradykinesia, quantified as a smaller and slower motion of the CoP, may be a primary deficit limiting forward postural stability in patients with PD. However, it is not known whether this type of postural bradykinesia has a similar effect on stability for all directions of disequilibrium in PD subjects. Our previous studies of PD subjects’ responses to multidirectional perturbations are consistent with the hypothesis that the basal ganglia are important for optimizing the pattern and magnitude of postural muscle synergies according to changes in the direction of perturbation or the size of the support base (Henry et al., 1998b, 2001). For each direction of postural sway, different sets of muscles are recruited to return the body to equilibrium (Henry et al., 1998b; Macpherson, 1998; Steiger et al., 1996; Weinrich et al., 1988). We found that PD subjects co-activate postural muscles in response to surface translations, resulting in abnormal direction of forces under each foot (Dimitrova et al., 2004a,b). Although the EMG deficits in PD subjects were not specific for particular directions of perturbations, it is unknown whether this co-contraction affects some directions of instability more than others. The current study investigated the postural stability and kinematic effects of this disordered muscle activation and horizontal surface forces during multidirectional postural perturbations to better understand the pathophysiology underlying postural instability in PD. It compared CoM, CoP, and kinematic hip and knee joint changes in response to 8 directions of surface translations and it investigated the ability of both subject groups to modify postural responses to maintain stability when their support base changed from wide to narrow stance.

Methods Subjects Seven healthy, elderly control and 7 patients with idiopathic PD (Hughes et al., 1992) were a subset of subjects included in studies of EMG responses and surface

506

F.B. Horak et al. / Experimental Neurology 193 (2005) 504–521

reactive forces (Dimitrova et al., 2004a,b). PD and control subjects with other causes of balance impairment (including somatosensory, visual, vestibular, or orthopedic disorders) were excluded. Patients with PD who had significant postural tremor, dysmetria, dystonia, or dementia were also excluded. All PD subjects had moderate to severe PD with significant balance problems (Table 1). Control subjects were healthy, ambulatory subjects recruited from the community. There were no significant differences in age (67 F 2 versus 66.3 F 2.2 years), height (167 F 5 versus 172 F 2 cm), weight (77 F 3 versus 76 F 8 kg), foot length (25 F 1 versus 25 F 1 cm), or foot width (9 F 0 versus 10 F 0 cm) between the control and PD groups, respectively. However, natural stance width between heels, measured after walking a few steps, was significantly narrower for the PD subjects (10.06 F 0.31 cm) than for the control subjects (19.0 F 1.16). All PD subjects responded to dopamine replacement therapy and were tested in the morning in the practical OFF state, after at least 12 h since their last dose. Immediately before each experiment, the PD subjects were scored with the Motor Subsection of the Unified Parkinson’s Disease Rating Scale (UPDRS) and the modified Hoehn and Yahr stages (Fahn and Elton, 1987). Table 1 summarizes the PD subjects’ age, duration of PD, clinical measures, and percent of fall trials during the experiment. All subjects gave informed consent according to the Declaration of Helsinki, and the Institutional Review Board of OHSU approved the study. Experimental protocol All subjects stood on dual force plates on a moveable platform, looking straight ahead with their arms at their sides. Subjects wore a harness attached to the ceiling without any tension, and an assistant stood at their left to assist in case of falls into the harness support. We excluded fall trials from analysis kinematics and kinetics although their EMG and initial force responses were analyzed in our previous studies (Dimitrova et al., 2004a,b).

Subjects were instructed to keep their balance without moving their arms or feet. Prior to each trial, subjects were coached to maintain the same initial force distribution under the feet, and during each trial, their initial force distribution was monitored on an oscilloscope by the experimenters. The platform was translated with 9-cm, 1000-ms duration (2 cm/ s2 peak acceleration) ramp and hold waveforms. Identical waveforms randomly translated the platform in 8 different directions. Platform translations were selected to allow most subjects to maintain their standing position without stepping. These platform parameters were slower than were those used in studies with young healthy subjects to allow some successful trials in the PD group (Henry et al., 2001). Figs. 1A and B show the CoM and CoP responses from a representative control subject to the 8 directions of passive body sway induced by platform translations in the opposite directions (F = forward, FR = forward-right, R = right, BR = backward-right, B = backward, BL = backward-left, L = left, FL = forward-left body sway). Throughout this paper, directions of postural perturbations will be referred to as the direction of passive body CoM and CoP displacement induced by surface translations in the opposite direction. Thus, in response to a forward-right perturbation, active postural responses (CoM and CoP) occur to bring the body back in a backward-left direction. A set of 8 randomized directions of translations was repeated 5 times, for a total of 40 trials administered in narrow stance and 40 trials in wide stance. Half of one subject group stood in wide stance first, and half of the other subject group stood in narrow stance first. Subjects rested between each set of 8 randomized directions of trials. Fig. 1C shows that, in narrow stance, parallel alignment of the inner side of the feet was separated by 4.5 cm and that, in wide stance, parallel foot alignment was separated by 26 cm. The size of the base of foot support at outer edges of feet was the same between the control subjects and PD subjects during the experiment, as shown in Fig. 1C (P N 0.05). As shown by the mean diameters of base of foot support for the control and PD subjects in Fig. 1C, the length of their base of foot support in narrow stance was

Table 1 Characteristics of subjects with Parkinson’s disease Subjects with PD

1 2 3 4 5 6 7 Mean SE

Age (years)

65 74 69 60 58 71 67 66.3 2.2

Duration of PD (years)

12 42 16 11 28 11 3 17.6 5.0

Total motor UPDRS

80.5 69 63 53 49 49 32 56.5 6.0

UPDRS subscores Item 28 Posture

Item 30 Postural Stability

Item 29 Gait

Item 22 Leg Rigidity

3 4 4 1 2 2 1.5 2.5 0.4

3 3 4 2 3 1 2 2.6 0.4

3 3 3 2 2 2 3 2.6 0.2

6 6 8 6 2 2 4 4.9 0.9

Hoehn and Yahr stage

% Falls in narrow stance

% Falls in wide stance

4 4 4 3 3 3.5 2.5 3.4 0.2

0 68 50 0 4 5 30 20.8 9.1

0 8 30 0 4 8 0 5.5 3.6

F.B. Horak et al. / Experimental Neurology 193 (2005) 504–521

507

Fig. 1. Representative examples of CoM (A) and CoP (B) displacements in the horizontal plane of a PD subject in response to 8 directions of body sway. The directions of body sway induced by platform translations as shown for the CoM displacement are the same for all polar plots in the following figures. Also shown in panel (C) are the mean size (cm) of support base for the 8 directions of perturbation in the PD and control subject groups and the distance between the feet in narrow stance and wide stance. Locations of reflective markers for kinematic and CoM analyses are shown in panel (D).

similar for the backward and lateral directions, which were both smaller than for the forward directions. In wide stance, the smallest size of the support base was still in the backward direction (11 cm), but the lateral and diagonal directions had a similar base of support (18–24 cm). Data collection and analysis Kinematics and center of mass A high-resolution Motion Analysis System (Santa Rosa, Ca) with 6 video cameras, sampling at 60 Hz, provided 3dimensional spatial coordinate information about body segment displacements. Reflective markers were placed bilaterally on the front of the body, at the center of joint rotation for the third metatarsophalangeal joints, ankles, knees, hips, shoulders, elbows, and wrists, as well as on the iliac crests, anterior–superior iliac spines, trunk, and head as used to calculate body CoM (Winter, 1979) (Fig. 1D). A reflective marker was also placed on the moving platform, and its displacement was subtracted from the body markers to measure body sway with reference to the surface. The

kinematic measurements were referenced to initial quiet standing. Twenty-six anthropometric measures of head, limb, and trunk segments (length, width, and perimeter) were collected from each subject (adapted from Chandler et al., 1975). These anthropometric data, together with the kinematic data, were used to calculate the position of CoM of each body segment in AP and lateral directions (Vaughan et al., 1991; Winter, 1979). The body CoM positions were derived as a weighted sum of all segments’ CoM positions (Vaughan et al., 1991). Kneejoint angle in the sagittal plane and hip-joint angle in the sagittal and frontal planes were calculated from adjacent segments. The peak joint-angle change over the duration of each trial was calculated from individual trials. Determined from individual trials were the peak CoM displacement from initial position, the time-to-peak CoM, initial rate of rise of CoM during the active postural response torques (from 150–850 ms following onset of surface translation), and the position of the CoM at the end of translation (1000 ms).

508

F.B. Horak et al. / Experimental Neurology 193 (2005) 504–521

Center of pressure Two force plates, each with 4 vertical and 2 horizontal strain gauge transducers, were mounted on the platform. The force signals were amplified and sampled at 120 Hz. Under each foot, AP CoP was derived from the difference between the 2 front and 2 back vertical forces and ML CoP from the difference between the 2 left and 2 right vertical forces. The vector sum of dtotal bodyT AP and ML CoP was used to quantify CoP responses, as shown in Fig. 1B. Total body AP CoP was calculated from the weighted sum of left and right foot CoP, by multiplying each foot’s CoP by the percent of total body weight on each foot (Henry et al., 1998b; Winter et al., 1996). Total body lateral CoP was calculated from the difference between the right and left foot plates vertical force, divided by the sum of vertical forces from the right and left plates, and summed with the lateral CoP under both feet (Henry et al., 1998a; Winter et al., 1996). The initial position of CoP in relation to the edges of foot support was measured after each experiment from tracings of tape outlining the feet (shown in Figs. 1C and D). From individual trials, we determined the peak CoP relative to the initial CoP position, time-to-peak CoP, the initial rate-of-rise of CoP from 150 to 350 ms following onset of surface translation, and the position of the CoP just before the end of translation (950 ms). The functional stability margin was calculated as the difference between the

peak CoP and peak CoM because the farther that the CoP is in front of the CoM, the less likely a subject is to fall in that direction (Winter, 1979). Statistical analysis To determine the effects of group (PD versus control), stance width (narrow versus wide), and platform displacement direction (8 directions), a 3-way ANOVA (1-between group and 2-repeated measures factors) were performed for each dependent variable: initial and maximal velocity of CoP displacement, peak CoP and CoM, time-to-peak CoP and CoM, stability margin (peak CoP–peak CoM), magnitudes of CoP and CoM at the time of platform offset, and peak hip- and knee-joint angle change. Post-hoc testing used the Newman–Keuls test.

Results CoP responses are smaller and slower in PD subjects than in control subjects Prior to platform translation, the initial position of CoP, with respect to the edges of foot support, were not significantly different between the PD subjects and control subjects (P N 0.05). In response to surface translations, the displacement of the CoP from initial position was a measure

Fig. 2. The mean CoP responses for control and PD subject groups over time in response to backward body sway and forward-diagonal body sway for (A) narrow stance and (B) wide stance. Onset and offset of platform translation are indicated by dashed lines at 0 and 1 s, respectively, in this and the following figures.

F.B. Horak et al. / Experimental Neurology 193 (2005) 504–521

of the forces used to return the body to equilibrium. In both stance widths, the PD subjects showed smaller peak CoP displacements compared to the control subjects (group effect in narrow stance: F [1,12] = 25, P b 0.001; in wide stance: F [1,12] = 12, P b 0.01). Fig. 2 illustrates group mean CoP displacements to backward body sway and to forwarddiagonal body sway in (A) narrow and (B) wide stance widths. Although the peak CoP displacement was smaller in PD subjects than in control subjects for all directions in narrow stance and 4 directions in wide stance, the peak CoP was not significantly different between the subject groups for lateral and forward diagonal directions in wide stance. For example, Fig. 2 shows larger CoP peak responses in the control group than in the PD group, in response to backward body sway in both stance widths and in response to forwardright body sway in narrow stance. However, as shown by the examples in Fig. 2, both subject groups displayed similar

509

CoP displacements in response to forward-right body sway in wide stance. Fig. 3A illustrates how the group mean (F SE) peak CoP (solid lines) was smaller for the PD group (black) than for the control group (gray) for each direction of body sway in narrow stance (left). In wide stance (right), the peak CoP differed for PD and control subjects for the 3 backwarddirected body sways (B, BL, and BR) as well as for the forward body sway. For example, for the PD group, backward perturbations in narrow stance resulted in a peak CoP of 5.9 F 0.2 cm compared to 6.8 F 0.2 cm for the control group. In wide stance, peak backward CoP displacement was 5.7 F 0.2 cm in the PD group compared to 6.8 F 0.2 cm in the control group. Note that the peak CoP was larger in response to lateral body sway than in response to forward or backward body sway in both groups during narrow stance but not in wide stance.

Fig. 3. A polar plot comparison of group mean (F SE) peak CoP (solid line) and CoM (dashed line) between (A) control (black) and PD subjects (gray) and (B) narrow and wide stance, in response to 8 directions of body sway. The shaded areas highlight the differences between control and PD group values (A) and between narrow and wide stance (B) in this figure and in Figs. 4, 5, and 7. Notice that the data points represent the actual position of peak CoP and CoM in the horizontal plane, and these positions are not always exactly opposite to the direction of platform translation. The directions of body sway and the scale (cm) are indicated in the top, left figure for all polar plots in this and the following figures.

510

F.B. Horak et al. / Experimental Neurology 193 (2005) 504–521

In both groups, the peak CoP displacement was larger in narrow, than in wide, stance (stance effect: in controls, F [1,12] = 18, P b 0.001; in PD, F [1,12] = 16.5, P b 0.002). Fig. 3B directly compares the mean CoP displacements (solid lines F SE) in narrow and wide stances for both subject groups. The ANOVA analysis showed that the effect of stance width on CoP displacement was greatly affected by the direction of perturbation (interaction: F [7,91] = 6.5, P b 0.001; compare Figs. 2A to B from single subjects). Peak CoP displacement was significantly larger in narrow, than in wide, stance for both groups for the lateral and forward diagonal directions of perturbations (Fig. 3B). For example, in response to rightward body perturbations, the peak CoP for PD subjects changed 1.1 cm from narrow to wide stance and for control subjects, it changed 1.8 cm. Peak CoP was not significantly different in both subject groups for narrow and wide stances in response to backward, backwarddiagonal, and forward body sway. In addition to the smaller amplitude of peak CoP displacements in the PD subjects compared to the control subjects, peak CoP also developed later in the PD subjects (F [1,12] = 11, P b 0.01; see Fig. 2 for examples). Across all directions in narrow stance, the peak CoP was 180 ms later for the PD subjects than for the control subjects. The timeto-peak CoP depended on the stance width and on the direction of perturbation (interaction: F [7,84] = 3.9, P b 0.001). In contrast to narrow stance, in wide stance the timeto-peak CoP was similar for PD subjects and control subjects in all directions, except for purely forward and backward falls, for which time-to-peak CoP was 170 ms later for the PD subjects than the control subjects. Altering stance width in the control group did not affect the timeto-peak CoP. In contrast, in the PD group, peak CoP displacements were significantly later in narrow, compared to wide, stance, especially for lateral perturbations (F [1,12] = 47, P b 0.001). In response to lateral perturbations, in narrow stance the peak CoP in PD subjects was 370 ms later than in wide stance. In conclusion, the time-to-peak CoP depended on the stance width and on the direction of perturbation (interaction: F [7,84] = 3.9, P b 0.001). In addition to the smaller and later peak CoP for the PD subjects, their initial rate-of-rise of CoP (150–350 ms after translation onset) was slower than that for the control subjects in both narrow and wide stance (in narrow stance: F [1,12] = 57, P b 0.001; in wide stance: F [1,12] = 16, P b 0.005). In Fig. 4A, the initial rate-of-rise of CoP is compared for PD and control group means (F SE), and in Fig. 4B, it is compared for narrow and wide stance in polar coordinates (see also the individual examples in Fig. 2). In narrow stance, CoP displacements were slower in the PD subjects for each direction of body sway, whereas in wide stance, the difference in CoP displacement between groups was significant only for the forward, backward, and backward-diagonal directions of body sway. Decrease of stance width (Fig. 4B) led to a faster CoP rate-of-rise in the control

Fig. 4. (A) Comparison of mean (F SE) initial rate-of-rise of CoP (150–300 ms after platform onset) in control versus PD subjects in narrow and wide stances over 8 directions of body sway. (B) Comparison of narrow versus wide stance in control and PD subjects for the same data shown in panel (A). The direction of body sway and the scale at the top left apply to all plots in this figure.

subjects but not in the PD subjects (control group: F [1,12] = 22, P b 0.001; PD group: F [1,12] = 2, P N 0.2). Not only did the control subjects move their CoP farther and faster than the PD subjects, but the control subjects also returned their CoP towards their initial stance position more quickly following platform perturbations. In contrast, the CoP of PD subjects remained displaced in the direction of body sway significantly longer (illustrated in the examples of backward body sway in Fig. 2). Just before perturbation offset (950 ms after platform onset), the CoP of control subjects was closer to their initial quiet stance position compared to the initial quiet stance position of PD subjects, for all directions in narrow stance (F [1,12] = 16, P b 0.005) and only for AP directions in wide stance. Across all directions, the control subjects returned their CoP to the same position in narrow and wide stances. In contrast, in narrow, but not wide, stance, the PD subjects kept their CoP significantly farther away from initial position following lateral and diagonal lateral perturbations. Peak CoM displacement is larger in PD subjects than in control subjects CoM displacement in response to surface translations was used as a measure of postural displacement from initial posture, with larger CoM displacements indicating greater instability. Fig. 3A compares the group mean peak CoM displacements (dashed lines) between PD (black) and

F.B. Horak et al. / Experimental Neurology 193 (2005) 504–521

control (gray) subject groups. In narrow stance, the PD subjects had larger CoM displacements than did the control subjects for all directions except for forward body sway (F [1,12] = 10, P b 0.01). In wide stance, however, CoM displacement was not significantly different for the PD and control groups. Like the CoP for both subject groups, the peak displacement of CoM was also larger in narrow, than in wide, stance (control subjects: F [1,12] = 13, P b 0.005; PD subjects: F [1,12] = 82, P b 0.001; Fig. 3B dashed lines). As expected, for pure AP directions of perturbation, peak CoM displacement was not affected by altering stance width (Fig. 3B). The control subjects showed more similarity in displacements of body CoM across directions and across stance widths (i.e., more circular polar plots) than did the PD subjects (Fig. 3B). The time-to-peak CoM was also delayed in the PD group, indicating a longer time to reverse CoM displacement (main group effect: F [1,12] = 6.7, P b 0.05). For time-to-peak CoM, there was a significant group  stance width  direction interaction (F [7,84] = 3.8, P b 0.001). On average, across all directions in narrow stance, the

511

beginning of the CoM displacement reversal was 220 ms later in the PD group compared to the control group. In wide stance, PD subjects also had later time-to-peak CoM than did control subjects for AP but not for lateral perturbations. The time-to-peak CoM displacement was similar in narrow and wide stances for the control subjects (F [1,12] = 0.1, P N 0.8), whereas for the PD subjects, the time-to-peak CoM displacement was significantly later in narrow, compared to wide, stance (F [1,12] = 26.3, P b 0.001). In all conditions for both subject groups, the mean time-to-peak CoM usually preceded the time of platform deceleration at 1000 ms, suggesting that subjects started to return their CoM to their initial equilibrium position prior to assistance from platform deceleration. However, in a small percent of PD subjects’ trials, the platform deceleration initiated the return of the CoM (Fig. 6, for examples). At the end of the perturbation, CoM displacement was significantly further away from initial position in the PD group than in the control group (main effect of group F [1,12] = 5.5;P b 0.05; see Fig. 5A for group means F SE). There was also a significant 3-way interaction

Fig. 5. (A) Comparison of mean (F SE) of CoM position at 950 ms after onset of platform translation (stopping at 1000 ms) in control and PD subjects in narrow and wide stances over 8 directions of body sway. (B) Comparison of narrow versus wide stance in control subjects and PD subjects for the data in panel (A).

512

F.B. Horak et al. / Experimental Neurology 193 (2005) 504–521

(F [7,84] = 4.0; P b 0.001) for CoM displacement at the end of the perturbation. In narrow stance, PD subjects were still leaning 4.9 F 0.03 cm (on average across directions of perturbations) whereas control subjects were only 3.3 F 0.01 cm from their initial CoM position when the platform decelerated to a stop (Fig. 5A, left polar plot). In contrast, in wide stance, there was no significant difference between PD and control subjectsT CoM displacement at the end of platform perturbation (Fig. 5A, right polar plot, P N 0.4). The effect of stance width on CoM displacement at the end of the perturbation was larger for PD subjects than for control subjects (Fig. 5B). The largest displacement of the CoM at the end of the perturbation (5.8 F 0.04 cm) was for PD subjects in narrow stance when responding to lateral perturbations (Fig. 5B, right polar plot). Stability margin is smaller in PD than control subjects The stability margin, defined by the difference between peak CoP and CoM displacement in each trial, was smaller in the PD subjects compared to the control subjects, with the stability margin depending on stance width and on the direction of perturbation (3-way interaction: F [7,84] = 3.7; P b 0.005). Fig. 6 shows examples of superimposed CoP and CoM trajectories

from a representative control subject and a representative PD subject during backward sway trials in narrow and wide stance. Notice the very small difference between peak CoP and CoM (shaded) in the PD subject compared to the control subject, in both narrow and wide stance for the backward body sway. Figs. 7A and B summarize the means and SEs of the stability margin for all 8 directions of body sway during narrow and wide stance in the PD and control subject groups. In narrow stance, the PD subjects showed significantly smaller stability margins than did the control subjects across all directions (F [1,12] = 24, P b 0.001). In wide stance, however, the PD subjects had smaller stability margins than did the control subjects only in response to forward, backward, and the 2 backward-diagonal body sways (Fig. 7A, right polar plot). In both groups, the stability margin was smallest for backward body sway. For the control subjects, the stability margin for backward body sway was 2.9 F 0.4 cm in narrow stance and 3.0 F 0.4 cm in wide stance, and for the PD subjects, it was only 1.1 F 0.4 cm in narrow stance and 1.0 F 0.6 cm in wide stance. Control subjects, but not PD subjects, adapted their responses to narrow and wide stance in ways that maintained their stability margin nearly constant across stance widths (control subjects: F [1,12] = 1; P N 0.4; PD subjects: Fig. 7B). The PD subjects had significantly smaller

Fig. 6. Examples of superimposed CoM and CoP displacements across time (s) during a backward sway displacement from a representative control subject (A) and a representative PD subject (#3 in Table 1) (B) in narrow and wide stance widths. Shaded areas and arrows illustrate that the CoP moves much more than the CoM in this control subject but not in this PD subject.

F.B. Horak et al. / Experimental Neurology 193 (2005) 504–521

513

Fig. 7. (A) Comparison of stability margin (difference between peak CoP and peak CoM displacements) in control versus PD subjects in narrow and wide stances over 8 directions of body sway (group mean F SE). (B) Comparison of stability margin in narrow versus wide stance in control and PD subjects for the data in panel (A). The direction of body sway and the scale at the top left apply to all plots in this figure.

stability margins in the narrow, compared to the wide, stance in the lateral perturbation directions (Fig. 7B, right polar plot; F [1,12] = 13; P b 0.005). PD subjects fall more than control subjects Control subjects never fell in response to platform translations, that is, they were never supported by the harness and were never caught by the attendant. In contrast, PD subjects fell in 13% of the trials, with falls occurring 4 times more often in narrow, than in wide, stance. However, the fall rates of the 7 PD subjects varied considerably, ranging from 0 to 57% of trials (see number of falls for each subject in Table 1). Most of their falls occurred in the backward sway direction (30% of backward sway trials) in both stance widths. Falls were next most common for the PD subjects in the narrow stance width, in response to the backward-diagonal and leftward sway directions (Table 1). Fig. 8 summarizes the distribu-

tion of fall incidence across the 8 directions of body sway in narrow and wide stance for PD subjects. For these subjects, the large number of backward falls corresponded to the smallest stability margin in the backward sway direction compared to the other sway directions (Fig. 7). The number of falls across directions was correlated with the group-average stability margin across directions (r = 0.60, P = 0.01). However, the number of falls across subjects did not significantly correlate with each subject’s average stability margin. In addition, for PD subjects, the number of falls did not correlate to certain clinical scores, such as scores on the backward-pull test (see Table 1), but their incidence of falls did correlate to their ages (r = 0.70, P b 0.05). Another possible reason for the PD subjects’ frequent falls is that they often remained leaning into their perturbed position until the offset of the perturbation (as shown in Fig. 6). The CoM redistribution was sometimes assisted by platform deceleration (1 s after the onset of perturbation).

514

F.B. Horak et al. / Experimental Neurology 193 (2005) 504–521

Fig. 8. Histogram showing the percent of trails in which PD subjects fell, across 8 directions of body sway in narrow (black) and wide (gray) stance widths.

However, for PD subjects, platform deceleration sometimes appeared to be a second perturbation. PD subjects often had a poor or absent response to surface deceleration, but control subjects produced a second CoP response to surface deceleration (Fig. 6). Hip and knee displacements are smaller in PD subjects than in control subjects Active hip and knee joint changes in response to the surface translations were smaller in PD subjects than in control subjects. Fig. 9A compares a time series of lateral hipangle changes in a representative PD subject and a representative control subject during left- and right-lateral surface translations, and Fig. 9B illustrates their knee flexion and extension during forward and backward surface translations. In the frontal plane, the feet are carried with the surface translation, temporarily leaving the trunk behind. During narrow stance, this passive hip displacement is followed by an active, lateral hip flexion so that the trunk is actively rotated in the same direction as passive body sway, to move the CoM back to equilibrium (Fig. 9A). The mean, active lateral flexion at the hip was 6 F 18 in the control group and 3 F 18 in the PD group. In the sagittal plane, immediately after the onset of surface perturbation, all subjects showed a small, passive knee flexion during forward body sway and a small, passive knee extension during backward body sway (Fig. 9B). These passive knee joint-angle changes are followed by active joint motion in the opposite direction. For example, active knee joint motion in response to backward body sway for the control versus PD groups were 6 F 18 versus 2 F 08 in narrow stance, and 6 F 28 versus 3 F 18 in wide stance, with a peak near the end of platform translation. Although the hip also flexed and extended in response to sagittal-plane translations, no significant differences were observed between subject groups.

Across all directions of sway, PD subjects had smaller peak displacements of hip and knee joints than did the control subjects (main effect of group: F [1,12] = 11; P b 0.01 for hip in frontal plane, F [1,12] = 6; P b 0.05 for knee in the sagittal plane). Figs. 9C and D compare the mean (F SE) peak hip displacements in the frontal plane and the peak knee displacements in the sagittal plane for both subject groups, as calculated from the right-sided joint markers (see Methods). In narrow stance, control subjects, compared to PD subjects, showed larger lateral trunk flexion to the right in response to rightward body sway, and larger lateral trunk flexion to the left in response to leftward body sway (Fig. 9C, left polar plot). In wide stance, lateral trunk flexion was very small in both subject groups (3 F 0.28 in the control group; 2 F 0.28 in the PD group; Fig. 9C, right polar plot). The lateral trunk flexion was significantly larger for narrow stance than in wide stance (main effect of stance: F [1,12] = 8, P b 0.05), although PD subjects did not significantly increase their lateral trunk flexion in narrow, compared to wide, stance. In both narrow and wide stances, control subjects also showed significantly larger knee flexion, particularly in response to backward and backward-diagonal body sway, than did PD subjects (Figs. 9D). Control subjects also showed more knee flexion in response to lateral translations in narrow, than in wide, stance (main effect of stance: F [1,12] = 10, P b 0.01). Unlike the control subjects, who broke their body motion into a multi-linked structure, PD subjects’ lack of active hip and knee motion in response to surface perturbations during both narrow and wide stance resulted in a stiff, inverted pendulum-type body sway. The clinical pull test is correlated with quantitative stability measures The backward pull test of the UPDRS (Postural Stability, Item 30) appears to be a good clinical indicator of the adequacy of automatic postural responses across PD subjects. The strongest correlations between the clinical measures of parkinsonism (Table 1) and quantitative stability data were between the pull test and (1) the stability margin in response to backward body sway (r = 0.78, P b 0.03); (2) the peak backward CoM (r = 0.75, P b 0.05); and (3) the CoM and CoP displacements at the end of translation (r = 0.78, P b 0.05; r = 0.94, P b 0.001). Fig. 10A shows the relationship between the pull test score and the backward stability margin, and Fig. 10B shows the relationship between the pull test score and the CoM position at the end of translation. The pull test was negatively correlated with the stability margin, suggesting that the worse the clinical score, the smaller the stability margin. One subject (#3) who had an average negative stability margin because he fell in so many trials also performed the worst in the pull test. In contrast, the pull test was positively correlated with the displacement of CoM at the end of translation, suggesting that the worse the clinical score, the further away from

F.B. Horak et al. / Experimental Neurology 193 (2005) 504–521

515

Fig. 9. Examples of changes in joint angles over time, in representative control and PD subjects for: (A) lateral motion of the hip in the frontal plane, in response to lateral perturbations, and (B) motion of the knee in the sagittal plane, in response to anterior and posterior perturbations. Shaded area indicates duration of surface translation. Large arrow (B) point to passive knee flexion and extension. Polar plots of peak flexion of joint angles (group mean F SE), in response to all 8 directions of body sway in narrow and wide stances are compared for the PD group (gray) and the control group (black) for (C) lateral flexion of the hip joint in the frontal plane and (D) flexion of the knee joint in the sagittal plane. Stick figures show the hip lateral flexion in response to a rightward body sway (leftward platform translation) and knee flexion in response to a backward body sway (forward surface translation). Both joint angles were measured from markers on the right side of the body (dots on stick figures). Hyperextension angles are omitted from polar plots.

516

F.B. Horak et al. / Experimental Neurology 193 (2005) 504–521

body sway. This outlier had the worst UPDRS score and a pull test score of 3, indicating severe retropulsion. Surprisingly, this subject showed a good stability margin (2.5 cm) during backward body sway. However, his large stability margin was not due to a large peak CoP (his peak CoP was the same as the mean peak CoP of the PD group). The good stability margin in this clinically unstable subject was due to a very small CoM displacement (e.g., 2.5 cm instead of the group mean of 5 cm). Thus, his rather large stability margin may have been due to unusually large passive stiffness from increased rigidity (UPDRS right + left leg rigidity = 6 out of 8) that limited the CoM displacement in response to the surface displacement (Horak et al., 1996b).

Discussion Direction-specific instability

Fig. 10. (A) Relationship between the stability margin (peak CoP–peak CoM) in response to backward body sway and clinical postural stability (UPDRS, pull test item #30) from the PD subjects, and (B) relationship between the backward displacement of CoM at the end of forward platform translation versus the score for clinical postural stability (UPDRS pull test item #30) in the PD subjects.

initial equilibrium position the CoM when the platform stops. This inability of the PD subjects to return their CoM to the initial position by the end of platform translation (at time 1 s) in both narrow and wide stances is illustrated in Fig. 6. The pull test was not significantly correlated with the initial CoP rate-of-rise or the peak CoP displacement, however. Furthermore, the quantitative measures were not significantly correlated with any other clinical scores listed in Table 1 (e.g., age, duration of PD, and UPDRS). As shown in Fig. 10A, the data from PD subject #1 showed an unusual relationship between his pull test score and his postural stability margin in response to backward

The postural stability margin, as defined by the difference between peak CoP and peak CoM, allowed us to define the relative postural stability for different directions of external perturbations as well as for PD and age-matched control subjects. Results from the current study show that control subjects demonstrated consistent postural stability margins for all 8 directions of postural perturbation. We had expected stability margins to vary across perturbation directions because of differences in biomechanical and musculoskeletal constraints on postural strategies for perturbations in different directions. For example, we had expected smaller stability margins for backward, than for forward, body sway because the foot cannot exert as much torque (and thus cannot displace the CoP) as much in the backward, than the forward, direction due to its short lever arm behind the ankle and relatively small dorsiflexor (compared to plantarflexor) muscles. However, the stability margin was the same for forward and backward perturbations in the control subjects. The nearly consistent stability margin for all directions of perturbation in control subjects suggests that the nervous system normally compensates for biomechanical differences across sway directions to maintain similar stability across directions of body sway. Specifically, to ensure similar stability margins for backward and forward body sway, the CNS must produce relatively larger dorsiflexor, than plantarflexor muscle activation as well as recruit more hip and trunk muscles for backward, compared to forward, body sway (Henry et al., 2001). Passive, biomechanical stability is much larger in the lateral than in the AP direction when the feet are far apart, so much lower levels of active muscle activation are required for postural responses to lateral perturbations than to AP perturbations during wide stance (Henry et al., 1998a, 2001). In contrast to age-matched control subjects, PD subjects showed different stability margins for different directions of body sway. The smallest stability margin in PD subjects occurred for backward body sway in both narrow and wide

F.B. Horak et al. / Experimental Neurology 193 (2005) 504–521

stance, suggesting that they are most vulnerable to falls in the backward direction, regardless of stance width (Fig. 7B, right). This observation explains why widening the base, a common recommendation for patients with postural instability, is unlikely to help the PD patient prevent backward falls. In addition, the widened base makes it more difficult for PD patients to initiate gait although it will increase the stability margin in the lateral direction (Gross et al., 2001). That PD subjects fell most often in response to backward body sway in both stance widths also supports the efficacy of using the CoP–CoM stability margin as a measure of dynamic postural stability. PD subjects’ small stability margin for backward body perturbations and their relative vulnerability to backward falls in the current study also supports the use of a backward tug to test PD patients’ postural stability in the clinic (Postural Stability, Item 30 of the UPDRS; Fahn and Elton, 1987). In fact, PD subjects’ results on the UPDRS’s Postural Stability Item significantly correlated with their backward stability margin, peak backward CoM, and peak backward CoM and CoP displacements at the end of translation. Unlike the backward tug test, which evaluates the ability to take a quick, compensatory step to regain balance, however, the current study quantified stability during bfeet-in-placeQ postural responses. The strong relationship between the quantitative postural measures in the current study and the clinical backward tug test is surprising because feet-in-place postural response strategies are thought to rely on different neural organization than compensatory stepping strategies (Horak and Macpherson, 1996; McIlroy and Maki, 1993). Although PD subjects were very vulnerable to falls in response to backward and backward-diagonal body perturbations for both stance widths, in narrow stance they also fell frequently in response to lateral sway. Both backward and lateral perturbation when the subject assumes a narrow stance require significant flexibility from the trunk (Henry et al., 1998a,b, 2001), so it is likely that axial rigidity and poor trunk coordination in PD subjects contributed to their poor stability in response to lateral and backward body sway. PD subjects were particularly unstable in response to lateral displacements during narrow stance because they had less lateral trunk flexion (Dimitrova et al., 2004b). In studies of young adults, an increase of active control of trunk muscles in response to lateral perturbations was found when they stood in narrow, compared to wide, stance (Henry et al., 1998a). Biomechanical models also predict that it is particularly important to add rapid trunk torques to limited leg torques to control postural equilibrium in a narrow base of foot support (Horak and Kuo, 2000; Kuo, 1995; Winter et al., 1990). These models show that CoM can be returned to equilibrium 3 times faster by flexing the hips and trunk than by using the body as a single inverted pendulum (Kuo and Zajac, 1993; McCollum and Leen, 1989). Whereas lateral CoP displacement for balance correction during wide stance is primarily accomplished by loading and unloading the 2 feet (Henry et al., 1998b, 2001; Winter et al., 1996), in

517

narrow stance, this strategy is insufficient to prevent the body’s CoM from reaching the limits of stability in response to external perturbations. Our previous study showed that, by rapidly flexing the trunk laterally, the control subjects created larger lateral shear forces at the surface, reflecting faster CoM lateral motion of the CoM, than did the PD subjects (Dimitrova et al., 2004a). Given their rigidity, our PD subjects could not rapidly reverse the faster and farther passive CoM displacement induced by perturbations. Instead, they barely maintained stability during surface translations and often kept from falling only by the deceleration of the support surface (Fig. 6B, right). PD subjects usually fell because their active CoP displacement was too slow and too small to reverse their falling CoM. PD subjects may lack trunk flexibility because of axial rigidity, as reflected in their co-contraction of hip and trunk muscles (Dimitrova et al., 2004a). Thus, they may be particularly unstable in response to lateral perturbations in narrow stance because of trunk stiffness. In response to lateral perturbations, we found that the control subjects, in fact, had more lateral hip flexion than did the PD subjects. Control subjects flexed the trunk in the direction of perturbation, allowing them to move their CoM more quickly back to equilibrium. In contrast, PD subjects held their bodies as rigidly inverted pendulums, which require larger surface reactive forces to return their CoM to equilibrium. Several studies have shown increased axial rigidity in PD subjects, especially in the OFF state (Schenkman et al., 1995; Steiger et al., 1996; Van Emmerik et al., 1999; Weinrich et al., 1988). Furthermore, rigidity at the knees or poor multi-joint coordination may have prevented the PD subjects from bending their knees in response to backward postural sway. Knee flexion helps actively move the CoM forward quickly, with less dorsiflexion torque required at the ankles (Crenna et al., 1987; Oddsson, 1990). We cannot explain the relative number of falls of our subjects across directions of perturbation solely on the size of the stability margin, especially with such a small subject group (Table 1). The narrow-stance stability margin for lateral perturbations was associated with frequent falls, yet this stability margin was about the same (about 2 cm) as the stability margin for forward body sway with no falls (Fig. 7A, left). When CoM displacement is expressed as a percent of the length of foot support, however, this normalized CoM displacement of PD subjects reflected their increased frequency of falls across different directions. For example, the 2 cm CoM displacement was 43–56% of support size for lateral perturbation but only 26% of support size for anterior perturbation. Although we do not know why PD subjects fell more to their left than their right, there are 3 potential explanations: (1) all PD subjects were right-handed, suggesting that they may have more strength in the right leg to resist perturbations to the right; (2) PD is usually asymmetrical, and 6 out of 8 of our subjects had worse

518

F.B. Horak et al. / Experimental Neurology 193 (2005) 504–521

Parkinsonian symptoms on their left side; and (3) an assistant stood on the subject’s left side, which might have reduced their fear of falling to the left. Postural bradykinesia In response to surface translations in various directions, CoP displacements and the net torque against the surface used for postural correction were significantly smaller and slower in the PD subjects, compared to age-matched control subjects. As a consequence of these slow, weak postural responses (bpostural bradykinesiaQ), PD subjects allowed their body CoM to be displaced further from their initial stance position and closer to the edges of their base of foot support. Thus, the dynamic stability margin (peak CoP– peak CoM) for PD subjects was approximately half that of the control subjects. Like the bradykinesia of triggered, automatic postural responses to surface translations, voluntary movements of upper and lower limbs are known to be weak in PD subjects (Kakinuma et al., 1998; Stelmach et al., 1989), and PD subjects are known to consistently undershoot their reaching to targets (Demirci et al., 1997). Voluntary force production impairments include slower rate-of-rise of force and more irregular force–time curves, similar to the PD subjects’ CoP curves observed in response to surface perturbations in the current study (Stelmach et al., 1989). However, when voluntary movements are triggered by external cues, movements can be as rapid and forceful as normal (Burleigh-Jacobs et al., 1997; Georgiou et al., 1993; Kelly et al., 2001; Morris et al., 1994). For example, the forces associated with anticipatory postural adjustments prior to voluntary step initiation can increase to normal levels when a voluntary step is triggered by external cues (BurleighJacobs et al., 1997). It has been hypothesized that the basal ganglia are particularly important for internally initiated movement, whereas extra-basal ganglia pathways control externally triggered movements (Brown and Marsden, 1988; Cunnington et al., 1995; Frank et al., 2000; Praamstra et al., 1998). Thus, it remains a mystery why automatic postural responses in response to surface translations are consistently weak and slow in PD subjects, although they are externally triggered. Parkinsonian bradykinesia of postural responses to surface translations is particularly puzzling because other types of externally triggered postural responses, such as postural responses prior to a compensatory stumble and responses to surface rotations, are not bradykinetic in PD subjects (Bloem et al., 1996; Gross et al., 2001). Although it is unlikely that automatic postural responses and voluntary movements result from the activation of the same basal ganglia–thalamic–cortical pathways (Horak and Macpherson, 1996; Massion, 1992), the similarity in bradykinesia between postural control and voluntary motor control suggests that the basal ganglia may play a similar role in benergizingQ both the ascending cortical motor routes for voluntary control as well as the descending motor routes

through the brainstem for automatic postural responses (Bejjani et al., 1997; Marsden, 1982). Like externally triggered voluntary movements, postural response latencies for PD subjects are normal, or slightly earlier-than-normal (Dimitrova et al., 2004a). However, the production of force in PD subjects is impaired in voluntary arm movements (Corcos et al., 1996; Stelmach et al., 1989;Vaillancourt et al., 2001) as well as in posture (Dietz et al., 1993) and gait (Abbs et al., 1987; Morris et al., 2001; Ueno et al., 1993), due to an abnormally low frequency of firing of motor units and due to greater motor unit discharge rate variability and impaired motor unit recruitment (Abbs et al., 1987). Another reason for PD subjects’ inability to generate force quickly for postural control is co-contraction. Previous studies show, co-activation of both distal, as well as proximal, muscles in response to multidirectional surface translations (Dimitrova et al., 2004a) or rotations (Carpenter et al., 2004). Similarly, PD subjects show excessive cocontraction upon the initiation of voluntary ankle movements (Hayashi et al., 1988). Thus, Parkinsonism is associated with bradykinesia of both voluntary and postural control. Not only did we find the peak CoP to be smaller than normal in our PD subjects, but the initial rate-of-rise of CoP was considerably slower in PD subjects than in the control subjects. The peak CoP in the PD subjects was 370 ms later than in control subjects, and the rate-of-rise of CoP in the PD subjects was 50% slower than that in the control subjects. This slower rate of change is likely due to cocontraction and poor motor unit recruitment, but may also be related to weaker muscles, either from atrophy or from a greater proportion of muscle fibers with slow twitch characteristics from years of tonic activation associated with rigidity (Dietz et al., 1993). However, studies of responses to surface rotations, instead of translations, have also shown reduced effective surface reactive torques due to larger, rather than smaller, muscle activations that are poorly coordinated (Carpenter et al., 2004). Adaptation to changing stance width Control subjects showed similar postural stability margins for both stance widths (Fig. 7B, left). These consistent postural stability margins suggest that the control subjects adaptively increased their postural response magnitude for lateral perturbations in narrow, compared to wide, stance to compensate for the reduced passive stabilization afforded by wider stance width. Control subjects similarly adapted the magnitude of their postural responses to backward, compared to forward, perturbations to achieve a consistent stability margin regardless of perturbation direction. The lateral stability margins of PD subjects and control subjects were similar in wide stance, but the PD subjects’ stability margin was much smaller than controls in narrow stance (Fig. 7B, right). Thus, PD subjects were unable to increase their weak postural responses to lateral perturba-

F.B. Horak et al. / Experimental Neurology 193 (2005) 504–521

tions when standing with a narrow, compared to a wide, base of support. However, the larger CoP displacement in PD subjects in wide, compared to narrow, stance was likely due to an increase in passive stiffness of their legs–pelvis complex rather than due to an increase in active postural response (Day and Steinger, 1993; Winter et al., 1996). Our previous study of surface reactive forces showed that passive reactive forces under each foot, prior to mediumlatency active postural responses, were larger than normal in PD subjects although their active forces were smaller than normal (Dimitrova et al., 2004b). Thus, passive stiffness and tonic rigidity in PD subjects may partially compensate for their weak postural muscle activation in response to perturbations. However, in response to large perturbations, passive stiffness and tonic rigidity cannot prevent falls because tonic stiffness impedes the generation of quick active movements, such as stepping, arm movements, and trunk flexion (Carpenter et al., 2004, McIlroy and Maki, 1993). In the current study, dwide stanceT at 26 cm between the heels is not actually very wide. It was smaller than our subjects’ average shoulder width and, similar to our elderly control subjects’ preferred stance width of 20 F 1.1 cm. However, the natural stance width for our PD subjects was considerably smaller at 10 F 0.3 cm. Subjects with severe PD and postural deficits are known to prefer a narrowerthan-normal stance width (Charlett et al., 1998; Nutt and Horak, 2002). Furthermore, stance width in gait and stance tends to become narrower the longer the duration of PD and the greater the subject’s rigidity (Charlett et al., 1998). We hypothesize that PD subjects prefer narrow stance width to allow adequate unloading of the swing leg for step initiation, given their weak anticipatory postural adjustments (Burleigh-Jacobs et al., 1997). However, our current study shows that this narrow stance severely compromises their lateral postural stability in response to perturbations. The inability of PD subjects to increase the magnitude of their active postural responses in response to lateral external displacements when standing with a narrow base of support may be due to the inability to use somatosensory information regarding initial stance configuration to adapt, shape, or tune postural response patterns so they are efficient for particular biomechanical conditions. This role of the basal ganglia in monitoring initial conditions to modify existing motor programs so they are specific for current contexts or constraints is consistent with the role of the basal ganglia in changing motor dsetT (Marsden, 1984). To flexibly modify postural responses for changes in biomechanical demands across directions of perturbation, the healthy nervous system needs accurate information from proprioceptors about perturbation characteristics and an accurate internal model of body biomechanics to accurately shape postural responses for consistent postural stability. The difficulty that PD subjects show in modifying their postural responses for biomechanical demands related to changes in direction and for changes in stance width may be due to poor use of

519

proprioceptive information (Zia et al., 2000) or to poorly developed internal representation of their body, necessary to customize motor programs (Horak and Macpherson, 1996). Why are PD subjects most unstable during backward and lateral body sway? This study, together with our previous publications with the same subjects, suggests that both biomechanical and neural constraints contribute to the directionally specific postural instability in PD (Dimitrova et al., 2004a,b). Backward sway and narrow based lateral sway are biomechanically more unstable because the CoM of the body comes closer to its limits of stability because the base of foot support is smaller and the ability of the ankles to exert torque are more limited in backwards and lateral directions than forwards (Horak and Macpherson, 1996). Rigid and bradykinetic postural responses, although not directionally specific in themselves (Dimitrova et al., 2004a), also result in directionally specific biomechanical constraints because recovery from backward sway demands knee flexion and recovery from lateral sway, with feet together, demands lateral trunk flexion, which are both limited by nonspecific stiffening responses in PD (Carpenter et al., 2004). In addition, the biomechanical effects of a flexed posture of PD subjects further limit the amount of torque they can generate at the support surface, exaggerating bradykinetic postural responses (Jacobs et al., 2004). A neural constraint specific to PD, the inability to flexibly modify postural responses based on postural demands may also contribute to directionally specific instability (Chong et al., 1999, 2000; Horak et al., 1992). This inflexibility can be seen as the inability to increase the rate of rise of CoP displacement for narrow stance and for backward sway. This direction specific modification of postural response magnitude and patterns is what allows postural stability margins to be similar across directions and stance widths in control, but not PD, subjects. We cannot be sure whether the inability to increase postural response magnitude for backward and lateral perturbations and for narrow stance is due to an inability to flexibly change postural set or to bradykinesia and rigidity because we only examined exposure to randomized directions and to wide, followed by narrow stance. Another neural constraint associated with PD, is poor motor control of axial muscles (Nutt and Horak, 2002). Recovery from backwards and lateral sway while in a narrow stance requires more control by proximal muscles and proximal muscles show more cocontraction and less directionally specific activation than distal muscles in PD subjects (Dimitrova et al., 2004a). Thus, these results suggest that the basal ganglia, in addition to regulating muscle tone and energizing muscle activation, also are critical for adapting postural response patterns for specific biomechanical conditions (Chong et al., 1999, 2000; Diener et al., 1987; Schieppati and Nardone, 1991). The basal ganglia appear to play a unique role in

520

F.B. Horak et al. / Experimental Neurology 193 (2005) 504–521

quickly modifying central motor (and cognitive) dsetT that allows immediate tuning of postural response patterns so they are specific for the particular situation (Horak and Macpherson, 1996). Thus, the complex interactions between biomechanical constraints on balance across different body sway directions and constraints on motor control imposed by basal ganglia pathology result in very specific patterns of falls in patients with Parkinson’s disease (Grimbergen et al., 2004).

Acknowledgments The authors thank Martine Mientjes for data analysis, Jill Knop for figure graphics, Leta Guptill and Patricia CarlsonKuhta for editorial assistance, and support from NIH grant AG-06457 (Dr. Horak) and HSFPO SF0011/1999-B (Dr. Dimitrova).

References Abbs, J.H., Hartman, D.E., Vishwanat, B., 1987. Orofacial motor control impairment in Parkinson’s disease. Neurology 37, 394 – 398. Allum, J.H., Carpenter, M.G., Honegger, F., Adkin, A.L., Bloem, B.R., 2002. Age-dependent variations in the directional sensitivity of balance corrections and compensatory arm movements in man. J. Physiol. (London) 542, 643 – 663. Bejjani, B., Damier, P., Arnulf, I., Bonnet, A.M., Vidailhet, M., Dormont, D., Pidoux, B., Cornu, P., Marsault, C., Agid, Y., 1997. Pallidal stimulation for Parkinson’s disease. Two targets? Neurology 49, 1564 – 1569. Bloem, B.R., Beckley, D.J., van Dijk, J.G., Zwinderman, A.H., Roos, R.A., 1996. Influence of dopaminergic medication on automatic postural responses and balance impairment in Parkinson’s disease. Mov. Disord. 11, 509 – 521. Brown, R.G., Marsden, C.D., 1988. Internal versus external cues and the control of attention in Parkinson’s disease. Brain 111, 323 – 345. Burleigh-Jacobs, A., Horak, F.B., Nutt, J.G., Obeso, J.A., 1997. Step initiation in parkinson’s disease: influence of levodopa and external sensory triggers. Mov. Disord. 12, 206 – 215. Carpenter, M.G., Allum, J.H.J., Honegger, A.L., Bloem, B.R., 2004. Postural abnormalities to multidirectional stance perturbations in Parkinson’s disease. J. Neurol. Neurosurg. Psychol. 75, 1245 – 1254. Chandler, R.F., Clauser, C.E., McConville, J.T., Reynolds, H.M., Young, J.W., 1975. Investigation of Inertial Properties of the Human Body. Aerospace Medical Research Laboratory, Washington, DC, pp. 1 – 162. Charlett, A., Weller, C., Purkiss, A.G., Dobbs, S.M., Dobbs, R.J., 1998. Breadth of base whilst walking: effect of ageing and Parkinsonism. Age Aging 27, 49 – 54. Chong, R.K.Y., Horak, F.B, Woollacott, M., 1999. Postural Set for Balance Control is Normal in Alzheimer’s But Not in Parkinson’s Disease. Chong, R.K.Y., Horak, F.B., Woollacott, M., 2000. Parkinson’s disease impairs the ability to change set quickly. J. Neurol. Sci. 175, 57 – 70. Corcos, D.M., Chen, C.-M., Quinn, N.P., McAuley, J., Rothwell, J.C., 1996. Strength in Parkinson’s disease: relationship to rate of force generation and clinical status. Ann. Neurol. 39, 79 – 88. Crenna, P., Frigo, C., Massion, J., Pedotti, A., 1987. Forward and backward axial synergies in man. Exp. Brain Res. 65, 538 – 548. Cunnington, R., Iansek, R., Bradshaw, J.L., Phillips, J.G., 1995. Movement-related potentials in Parkinson’s disease: presence and predictability of temporal and spatial cues. Brain 118, 935 – 950.

Day, B., Steinger, M.J., 1993. Effect of vision and stance width on human body motion with standing: implications for afferent control of lateral sway. J. Physiol. 469, 479 – 499. Demirci, M., Scholz, E., McShane, L., Hallett, M., 1997. A mismatch between kinesthetic and visual perception in Parkinson’s disease. Ann. Neurol. 41, 781 – 788. Diener, H.C., Scholz, E., Guschlbauer, B., Dichgans, J., 1987. Increased shortening reaction in Parkinson’s disease reflects a difficulty in modulating long loop reflexes. Mov. Disord. 2, 31 – 36. Dietz, V., Zijlstra, W., Assainte, C., Trippel, M., Berger, W., 1993. Balance control in Parkinson’s disease. Gait Posture 1, 77 – 84. Dimitrova, D., Horak, F.B., Nutt, J.G., 2004a. Postural muscle responses to multidirectional translation in patients with Parkinson’s disease. J. Neurophysiol. 91, 489 – 501. Dimitrova, D., Horak, F.B., Nutt, J.G., 2004b. Abnormal force patterns for multidirectional postural responses in patients with Parkinson’s disease. Exp. Brain Res. 156, 183 – 195. Fahn, S., Elton, R., 1987. Unified Parkinson’s disease rating scale. In: Fahn, S., Marsden, D., Calne, D. (Eds.), Recent Developments in Parkinson Diseases. Macmillan, London, pp. 153 – 163. Frank, J.S., Horak, F.B., Nutt, J.G., 2000. Centrally initiated postural adjustments in Parkinsonian patients on and off levodopa. J. Neurophysiol. 84, 2440 – 2448. Georgiou, N., Iansek, R., Bradshaw, J.L., Phillips, J.G., Mattingley, J.B., Bradshaw, J.A., 1993. An evaluation of the role of internal cues in the pathogenesis of Parkinsonian hypokinesia. Brain 116, 1575 – 1587. Greenspan, S.L., Myers, E.R., Kiel, D.B., et al., 1998. Fall direction, bone mineral density, and function: risk factors for hip fracture in frail nursing home elderly. Am. J. Med. 104, 539 – 545. Grimbergen, Y., Munneke, M., Bloem, B.R., 2004. Falls in Parkinson’s disease. Curr. Opin. Neurol. 17, 405 – 415. Gross, A., Dimitrova, D., Horak, F.B., 2001. The effects of stance width on automatic postural responses in Parkinson’s disease. International Society for Posture and Gait. Maastricht, The Netherlands. Hayashi, A., Kagamihara, Y., Nakajima, Y., Narabyshi, H., Okuma, Y., Tanaka, R., 1988. Disorder in reciprocal innervation upon initiation of voluntary movement in patients with Parkinson’s disease. Exp. Brain Res. 70, 437 – 440. Henry, S.M., Fung, J., Horak, F.B., 1998a. Control of stance during lateral and anterior/posterior surface translations. IEEE Trans. Rehabil. Eng. 6, 32 – 42. Henry, S.M., Fung, J., Horak, F.B., 1998b. EMG responses to maintain stance during multidirectional surface translations. J. Neurophysiol. 80, 1939 – 1950. Henry, S.M., Fung, J., Horak, F.B., 2001. The effect of stance width on multidirectional postural responses. J. Neurophysiol. 85, 559 – 570. Horak, F.B., Kuo, A., 2000. Postural adaptation for altered environments, tasks, and intentions. In: Winters, J.M., Crago, P.E. (Eds.), Biomechanics and Neuronal Control of Posture and Movement. Springer, New York, pp. 267 – 281. Horak, F.B., Macpherson, J.M., 1996. Postural orientation and equilibrium. In: Smith, J.L. (Ed.), Handbook of Physiology: Section 12: Exercise: Regulation and Integration of Multiple Systems, vol. 12. Oxford Univ. Press, New York, pp. 255 – 292. Horak, F.B., Nutt, J.G., Nashner, L.M., 1992. Postural inflexibility in Parkinsonian patients. J. Neurol. Sci. 111, 46 – 58. Horak, F.B., Frank, J.S., Nutt, J.G., 1996a. Effects of dopamine on postural control in Parkinsonian subjects: scaling, set and tone. J. Neurophysiol. 75, 2380 – 2396 (MD). Horak, F.B., Frank, J.S., Nutt, J.G., 1996b. Effects of dopamine on postural control in Parkinsonian subjects: scaling, set, and tone. J. Neurophysiol., 75, 2380 – 2396. Hughes, A.J., Daniel, S.E., Kilford, L., Lees, A.J., 1992. Accuracy of clinical diagnosis of idiopathic Parkinson’s disease: a clinico– pathological study of 100 cases. J. Neurol., Neurosurg. Psychiatry 55, 181 – 184.

F.B. Horak et al. / Experimental Neurology 193 (2005) 504–521 Jacobs, J., Dimitrova, D., Nutt, J., Horak, F., 2004. Does the stooped posture of Parkinson’s disease compensate for or contribute to abnormal postural responses? Mov. Disord. 19 (S9), S169 – S170 (Suppl.). Kakinuma, S., Nogaki, H., Pramanik, B., Morimatsu, M., 1998. Muscle weakness in Parkinson’s disease: isokinetic study of the lower limbs. Eur. Neurol. 39, 218 – 222. Kelly, V.E., Hyngstrom, A.J., Rundle, M.M., Bastian, A.J., 2001. Interaction of levodopa and cues on voluntary reaching in Parkinson’s disease. Mov. Disord. J. 12. Kuo, A., 1995. An optimal control model for analyzing human postural balance. IEEE Trans. Biomed. Eng. 42, 87 – 101. Kuo, A.D., Zajac, F.E., 1993. A biomechanical analysis of muscle strength as a limiting factor in standing posture. J. Biomech. 26, 137 – 150 (Elmsford NY). Macpherson, J.M., 1998. Strategies that simplify the control of quadrupedal stance II. Electromyographic activity. J. Neurophysiol. 60, 218 – 231 (MD). Macpherson, J.M., Horak, F.B., Dunbar, D.C., 1988. Stance dependence of automatic postural adjustments in humans. Exp. Brain Res. 78, 557 – 566. Marsden, C.D., 1982. The mysterious motor function of the basal ganglia: the Robert Wartenberg lecture. Neurology 32, 514 – 539. Marsden, C.D., 1984. Which motor disorder in Parkinson’s disease indicates the true motor function of the basal ganglia? Functions of the Basal Ganglia. Pitman, London, pp. 225 – 250. Massion, J., 1992. Movement, posture and equilibrium: interaction and coordination. Prog. Neurobiol. 38, 35 – 56. McCollum, G., Leen, T.L., 1989. Form and exploration of mechanical stability limits in erect stance. J. Mot. Behav. 21, 225 – 244. McIlroy, W.E., Maki, B.E., 1993. Changes in early dautomaticT postural responses associated with the prior-planning and execution of a compensatory step. Brain Res. 631, 203 – 211. Morris, M., Iansek, R., Matyas, T.A., Summers, J.J., 1994. The pathogenesis of gait hypokinesia in Parkinson’s disease. Brain 117, 1169 – 1181. Morris, M.E., Huxham, F., McGinley, J., Dodd, K., Iansek, R., 2001. The biomechanics and motor control of gait in Parkinson disease. Clin. Biomech. 16, 459 – 470. Nashner, L.M., Cordo, P.J., 1981. Relation of automatic postural responses and reaction-time voluntary movements of human leg muscles. Exp. Brain Res. 43, 395 – 405. Nutt, J.G., Horak, F.B., 2002. Gait and balance disorders. In: Asbury, A.K., McDonald, W.I., McArthur, J.C., McKhann, G., Goadsky, P. (Eds.), Diseases of the Nervous System: Clinical Neuroscience and Therapeutic Principles, 3rd, vol. 1. University Press Cambridge, Cambridge, pp. 581 – 591.

521

Oddsson, L.I., 1990. Control of voluntary trunk movements in man: mechanisms for postural equilibrium during standing. Acta Physiol. Scand. Praamstra, P., Stegeman, D.F., Cools, A.R., Horstink, M.W., 1998. Reliance on external cues for movement initiation in Parkinson’s disease evidence from movement-related potentials. Brain 121, 167 – 177. Schenkman, M., Hughes, M.A., Bowden, M.G., Studenski, S.A., 1995. A clinical tool for measure functional axial rotation. Phys. Ther. 75, 151 – 156. Schiepatti, M., Nardone, A., 1991. Free and supported stance in Parkinson’s disease: the effect of posture and dpostural setT on leg muscle responses to perturbation and its relation to the severity of the disease. Brain 114, 1227 – 1244. Steiger, M.J., Thompson, P.D., Marsden, C.D., 1996. Disordered axial movement in Parkinson’s disease. J. Neurol., Neurosurg. Psychiatry 61, 645 – 648. Stelmach, G.E., Teasdale, N., Phillips, J., Worringham, C.J., 1989. Force production characteristics in Parkinson’s disease. Exp. Brain Res. 76, 165 – 172. Ueno, E., Yanagisawa, N., Takami, M., 1993. Gait disorders in Parkinsonism: a study with floor reaction forces and EMG. Adv. Neurol. 60, 414 – 418 (NY). Vaillancourt, D.E., Slifkin, A.B., Newell, K.M., 2001. Visual control of isometric force output in Parkinson’s disease. School of Kinesiology. University of Chicago, Chicago, IL, p. 1. Van Emmerik, R.E., Wagenaar, R.C., Winogrodzka, A., Wolters, E.C., 1999. Identification of axial rigidity during locomotion in Parkinson disease. Arch. Phys. Med. Rehabil. 80, 186 – 191. Vaughan, C.L., Davis, B.L., O’Connor, J.C., 1991. Dynamics of Human Gait. Human Kinetics Publishers, Champaign, IL. Weinrich, M., Koch, K., Garcia, F., Angel, R.W., 1988. Axial versus distal motor impairment in Parkinson’s disease. Neurology 38, 540 – 545. Winter, D.A., 1979. Biomechanics of Human Movement. John Wiley and Sons, New York. Winter, D.A., Patla, A.E., Frank, J.S., 1990. Assessment of balance control in humans. Med. Prog. Technol. 16, 31 – 51 (MA). Winter, D.A., Prince, F., Frank, J., Powell, C., Zabjek, K., 1996. Unified theory regarding A/P and M/L balance in quiet stance. J. Neurophysiol. 75, 2334 – 2343 (MD). Yang, J.F., Winter, D.A., Wells, R.P., 1990. Postural dynamics in the standing human. Biol. Cybern., vol. 62, pp. 309 – 320. Berlin. Zia, S., Cody, F.W., O’Boyle, D.J., 2000. Joint position sense is impaired by Parkinson’s disease. Ann. Neurol. 47, 218 – 228.