INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The cerebellum is important for human limb coordination (Bastian et al. 1996; Gilman et al. 1976; Goodkin et al. 1993), control of upright stance and balance (Diener et al. 1984; Horak and Diener 1994; Mauritz et al. 1979), and locomotion (Earhart and Bastian 2001; Palliyath et al. 1998). Locomotor abnormalities often occur after cerebellar damage in humans and are characterized by a veering, stumbling path, wide base of support, impaired multi-joint coordination, decomposition of joint movements in the leg, and irregular and more variable foot placement (Crowdy et al. 2000; Earhart and Bastian 2001; Hallett and Massaquoi 1993; Palliyath et al. 1998). These features of cerebellar gait ataxia could be associated with a primary impairment of balance, of limb coordination, or of some combination of the two.

Physiological studies of monkeys and cats provide evidence for a functional localization within the cerebellum, with control of posture, equilibrium, and locomotion relatively localized to the medial zone (vermis and fastigial nuclei), control of discrete, ipsilateral limb movements and reflexes relatively localized to the intermediate zone (intermediate hemisphere and interpositus nuclei), and control of complex, visually guided limb movements and the planning of those movements relatively localized to the lateral zone (lateral hemisphere and dentate nuclei) (Arshavsky et al. 1983; Botterell and Fulton 1938a,b; Chambers and Sprague 1955a,b; Orlovskii 1972; Yu and Eidelberg 1983). These physiological findings are in agreement with anatomical evidence showing somewhat isolated neuronal connections to and from these functional zones in the cerebellum (Brooks and Thach 1981; Voogd and Glickstein 1998). The medial zone of the cerebellum receives afferent information from primary visual, auditory, vestibular, and somatosensory structures and projects mainly back to the brain stem (e.g., vestibular and reticular nuclei) and a bit more sparsely to the motor cortical regions via thalamus (Asanuma et al. 1983c). The intermediate zone receives most of its afferent information from somatosensory receptors from the limbs and projects both to motor cortical regions via thalamus and to brainstem regions (Brooks and Thach 1981). The lateral zone of the cerebellum receives from motor, premotor, and prefrontal cortical regions and projects back to them via thalamus (Asanuma et al. 1983a,b; Middleton and Strick 2001). Thach and colleagues (1992) found that temporary inactivation of the fastigius tended to produce deficits of sitting and standing balance, falls to the side of the lesion and abnormal locomotion but left isolated limb movements comparatively normal; inactivation of the dentate nucleus tended to produce deficits of reach and grasp but left locomotion relatively intact. These animal studies suggest that cerebellar control of balance and locomotion may be relatively localized to the medial zone and that locomotor abnormalities may be more closely associated with deficits of balance than with deficits of voluntary, visually guided limb movements.

Other recent evidence, however, suggests that the lateral region of the cerebellum may also play a role in certain types of locomotion. In cats, neural activity in the dentate nucleus is related to unexpected perturbations during treadmill locomotion (Schwartz et al. 1987). Purkinje cells in the lateral cerebellar hemisphere increase firing during walking on a horizontal, circular ladder (Marple-Horvat and Criado 1999) and during adaptations to elevated ladder rungs (Armstrong and Marple-Horvat 1996; Armstrong et al. 1997; Marple-Horvat et al. 1998). Thus the lateral regions of the cerebellum may help to control locomotion, particularly when adjustment of the motor output to novel contexts and/or strong visual guidance is required.

Cerebellar functional localization for the control of human locomotion is less well understood. Several studies have shown that human cerebellar damage produces abnormalities of locomotion (Crowdy et al. 2000; Earhart and Bastian 2001; Hallett and Massaquoi 1993; Palliyath et al. 1998) that resemble those described in animals. Hallett and Massaquoi (1993) found that individuals with cerebellar damage lacked a consistent step direction and distance, had a "drifting" or veering path of movement, exhibited more variable joint angle excursions, and had impaired coordination between the ankle and knee joints. Another study showed that subjects with cerebellar damage lacked foot-placement accuracy when using visual guidance to step on targets during walking (Crowdy et al. 2000). Still, no study has related human cerebellar walking abnormalities to a particular type of motor impairment, such as a balance or leg-coordination deficit.

Given the similar patterns of gait ataxia across animal species, it may be that the physiological mechanisms for the control of locomotion in humans are similar to those of other mammals. Thus we might predict that cerebellar deficits during uninterrupted, level-ground walking would be more related to balance disturbances than leg-coordination deficits. However, the fact that human locomotion is bipedal may introduce some differences. Bipedal locomotion is less stable than quadrupedal locomotion and likely requires additional descending cerebral cortical control. It is therefore possible that bipedal locomotion requires additional contributions from the lateral cerebellum (which has heavy projections to and from cerebral cortical regions) that are less critical in quadrupedal locomotion. If this was the case, cerebellar deficits during human locomotion might also be associated with voluntary, visually guided leg-coordination deficits.

One difficulty in studying functional localization within the human cerebellum is that the damage is often not restricted to a single functional zone. Stroke and degenerative diseases are common causes of cerebellar damage and typically affect multiple zones (Amarenco 1991; Gomez et al. 1997; Tatu et al. 1996). However, animal studies suggest that deficits of balance and voluntary, visually guided limb movement are dissociable (Botterell and Fulton 1938a,b; Chambers and Sprague 1955a,b; Thach et al. 1992). Therefore we directed our study at examining the relative contributions of these two types of motor deficits to cerebellar gait ataxia. The purpose of this study was to determine whether specific features of cerebellar gait ataxia, observable during uninterrupted walking over level ground, could be attributed to either a balance deficit or a voluntary leg-coordination deficit. We predicted that certain features of human cerebellar gait ataxia would be associated with deficits of balance, whereas others would be more closely associated with deficits of voluntary leg coordination. We hypothesized that cerebellar subjects with a balance deficit would walk with a wide stance and stiff-legged pattern characterized by decreased stride lengths, increased stride widths, and decreased peak joint angles. We predicted that cerebellar subjects with a leg-placement deficit would walk with a dysmetric, high-stepping pattern characterized by increased stride-length variability, increased foot-path curvature, and increased joint decomposition. Because we considered speed to be a rather nonspecific measure of walking impairment and because reducing speed might be an effective compensatory strategy to minimize the effects of a balance or leg-placement deficit, we predicted that cerebellar subjects with either type of deficit would demonstrate reduced walking speeds. We found that balance deficits related to many features of cerebellar gait ataxia, but leg-placement deficits did not. Preliminary results of this work have been previously presented (Morton and Bastian 2001).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Twenty subjects with cerebellar damage [6 females and 14 males; mean age 46.40 ± 4.85 (SE) yr] and 20 age- and gender-matched healthy control subjects (45.90 ± 4.67 yr) participated in the study. All subjects gave their informed consent prior to participating, and a Human Studies Committee approved the study. Cerebellar damage was confirmed by MRI or computed tomography (CT) scan, and localization to a specific zone was determined when possible. Prior to testing, all cerebellar subjects underwent a thorough motor neurological examination. Subjects with clinical evidence of involvement of other brain structures (e.g., motor weakness, sensory loss, hyperreflexia, bradykinesia, rigidity) were excluded from the study. See Table 1 for cerebellar subject information. Additional detailed information regarding diagnosis and lesion location (when available) is provided in the following text.

Twelve of the cerebellar subjects (CBL-1, -2, -3, -4, -6, -7, -11, -12, -13, -14, -16, and -17) had degenerative diseases. In all of these subjects, cerebellar damage was diffuse and we were unable to isolate the lesion to a specific zone. Five of the cerebellar subjects with degenerative diseases tested positive for a specific genetic spinocerebellar ataxia. Subjects CBL-2, -11, -12, and -17 were diagnosed with SCA 6. Subject CBL-1 was diagnosed with SCA 3. SCA 6 tends to be a relatively purely cerebellar disease, whereas SCA 3 is not. However, both the MRI report and the neurological examination from subject CBL-1 failed to show any sign of extracerebellar involvement. The other seven subjects with cerebellar degeneration (CBL-3, -4, -6, -7, -13, -14, and -16) were diagnosed with idiopathic pancerebellar atrophy. On neurological examination, none of these subjects had evidence of involvement of any brain region beyond the cerebellum, and none had any signs of autonomic system dysfunction.

The other eight cerebellar subjects had more focal lesions. Five subjects (CBL-9, -15, -18, -19, and -20) all had a vermal split. The cerebellum is often partially split through the posterior vermis for removal of tumors. Subject CBL-15 had a surgical split of the vermis for removal of an astrocytoma. Subject CBL-9 had a surgical split of the superior vermis, and subjects CBL-18, -19, and -20 had surgical split of the posterior vermis, all for removal of a medulloblastoma. Subject CBL-15 had postsurgical complications that resulted in some damage to the brainstem. Subjects CBL-18 and -19 underwent postsurgical chemotherapy and radiation. The other three subjects with more focal lesions were subjects CBL-5, 8, and -10. Subject CBL-5 had a hemorrhage. The CT scan indicated that the hemorrhage involved the cerebellar vermis with slight edema to nearby cerebellar regions. Subject CBL-10 also had a vermal hemorrhage. The MRI report indicated that the hemorrhage was localized to the vermis, again with a small amount of edema detected in nearby areas of the cerebellum. Subject CBL-8 had bilateral superior cerebellar artery strokes.

Paradigm

Prior to testing, we rated the severity of ataxia using the International Cooperative Ataxia Rating Scale (ICARS), a 100-point ordinal scale measure that quantifies ataxia in four categories of movement: posture and gait, limb kinetics, speech, and eye movements (Trouillas et al. 1997). Higher scores indicate increased impairment. We measured ICARS scores in 18 of the 20 cerebellar subjects.

All subjects performed three sets of tasks: a balance task, a leg-placement task, and walking. The balance task was a dynamic, voluntary lateral weight-shifting task. The leg-placement task was a voluntary, visually guided stepping movement. Based on performance during these two relatively isolated tasks, each cerebellar subject was placed into one of four subgroups: no detectable deficits, balance deficit only, leg-placement deficit only, or both balance and leg-placement deficits. We then determined the walking abnormalities associated with each of the cerebellar subgroups. Detailed information about each of the tasks is given in the following text.

Balance task

Subjects stood on a force plate with eyes open and feet shoulder-width apart. Arms were held across the chest. Subjects were instructed to repeatedly shift their weight back and forth laterally as far as possible without picking up their feet. Data were recorded for 20-s intervals for each trial. This task was specifically designed to closely resemble the balance requirements for normal walking. During the late stance phase of gait, the center of mass is shifted laterally over one limb, to free the other limb from support so that it may be advanced forward during swing phase.

Leg-placement task

Subjects made single steps onto a visual target, which was a tracing of the foot located on the floor. At the beginning of each trial, subjects' feet were positioned approximately shoulder-width apart with the toe of the stepping leg parallel to the heel of the stationary leg. The target was located directly in front of the stepping leg, at a distance of approximately 1-1.5-foot lengths from the stationary leg. Subjects were asked to step as accurately as possible onto the target without hesitating or correcting after landing. All subjects performed the task while holding onto parallel bars with both hands, thus reducing the effects that a balance deficit may have had on performance. Even light fingertip contact with a support surface has been shown to reduce postural sway in healthy subjects (Holden et al. 1994) and in individuals with vestibular loss (Lackner et al. 1999). To verify that the support bars provided sufficient stability during this task, we compared angular excursions of the trunk in the anterior-posterior and medial-lateral dimensions. Trunk angular excursions were minimal and did not differ between groups, and trunk excursions were not correlated with performance of the task. Further, we found that even subjects with profound balance deficits could perform this task normally (e.g., CBL-5). The leg-placement task was specifically designed to closely match the stereotypical joint kinematics and foot trajectory of the swing limb during normal walking, without the balance requirements.

Walking task

The walking task consisted of walking as fast as possible, across a level, unobstructed walkway, approximately 8 m in length. Subjects who could not walk alone safely (3 of 20 cerebellar subjects; CBL-6, -11, and -12) were given handhold assistance by an examiner.

Subjects received practice with the balance and leg-placement tasks prior to recording data. Subjects performed two to four trials of the balance task and four to eight trials of the leg-placement task and the fast-paced walk. An examiner guarded subjects during all tasks and all subjects received rest breaks as needed throughout the testing session to minimize fatigue.

Data collection

Joint positions were recorded in three dimensions using the OPTOTRAK System (Northern Digital, Waterloo, ON). Six infrared light-emitting diodes (IREDs) were placed on one of the legs and the trunk, marking the positions of the foot (head of fifth metatarsal), ankle (lateral malleolus), knee (lateral knee joint space), hip (greater trochanter), pelvis (superior iliac crest), and shoulder (head of humerus; see Fig. 1). Three additional IREDs were placed on the contralateral leg (foot, heel, and ankle) to monitor the location of each foot relative to the other. We tested the more involved leg of the subjects with cerebellar damage and the corresponding leg of the control subjects, matched for dominance. If, on neurological examination, neither leg appeared to be more involved, we tested the dominant leg. Subjects were asked which leg they would prefer to kick a ball with to determine leg dominance. During the balance task, we also recorded center of pressure coordinates using an AMTI LG6-2-1 force plate. We defined the coordinates of our laboratory space such that anterior-posterior movements were in the x direction, vertical movements in the y direction, and medial-lateral movements in the z direction. Position data were collected at 100 Hz and force plate data at 1,000 Hz.

View larger version (19K): [in this window] [in a new window]   Fig. 1. Experimental setup. Subjects stood on the force plate with arms across the chest and feet shoulder-width apart during the balance task. Subjects held onto parallel bars during the leg-placement task. Subjects walked across an uninterrupted, level platform during the walking task.

Data analysis

Joint position and force plate data were low-pass filtered at 10 Hz. OPTOTRAK software was used to calculate marker positions, velocities, and joint angles and to generate animated stick figures that were used to identify the times of heel strike and toe off during the fast-paced walk. Custom software from Matlab (Mathworks) was used for all other analyses.

We quantified balance deficits by measuring the lateral excursion of the center of pressure during the dynamic weight-shifting task. For each trial, the lateral movement of the center of pressure was plotted versus time. We then identified peaks in the path of the center of pressure and calculated the peak-to-peak lateral (z) distance traveled by the center of pressure during each shift in weight. Weight-shifting distances were averaged over all shifts and all trials and normalized to the spread of the feet, measured as the z distance between the second metatarsals.

We quantified leg-placement deficits by measuring endpoint errors in the foot location. We used the ankle tangential velocity trace to identify the times of start and stop of movement. Start of movement was defined as the first time the velocity exceeded 5% of the peak; end of movement was defined as the time when the velocity dropped less than 5% of the peak and the foot was on the floor. Foot endpoint errors were calculated as the vector distance between the foot marker location at the end of movement and the foot marker location when ideally placed on the target. Foot endpoint errors were calculated and averaged over all trials for each subject.

We quantified walking performance with a number of variables specifically selected to address known features of cerebellar gait ataxia. We measured stride length, stride-length variability, stride width, peak joint angles, foot-path curvature, joint-joint decomposition indices, and walking speed. For each trial, we first identified the times of heel strike and toe off using the animated stick figures. We defined a stride as the period from heel strike to heel strike, which encompassed stance (heel strike to toe off) and swing (toe off to heel strike) phases. Stride length was measured as the forward (x) distance traveled by the ankle marker, from one heel strike to the next, normalized to subject height. We quantified trial-to-trial variability in stride length with a coefficient of variation (SD in stride length/mean stride length × 100). Lower coefficients of variation indicated less trial-to-trial variability in the placement of the foot. Stride width was measured as the lateral (z) distance between the right and left ankle markers at the time of heel strike, normalized to subject height. Peak joint angles were calculated at the hip, knee, and ankle joints. Specifically, we measured peak hip flexion, peak knee flexion and peak ankle plantarflexion. Foot-path curvature was measured as the ratio of the three-dimensional path to the straight line path of the foot marker during swing. A foot-path ratio of 100% indicated a perfectly straight trajectory, whereas higher values indicated increased curvature of the foot path. We also calculated joint-joint decomposition indices. Decomposition of movement is a pattern first described by Holmes (1939) that reflects the tendency for people with cerebellar ataxia to move only one joint at a time. To quantify this tendency, we generated decomposition indices at each of the major joint pairs of the leg. The decomposition index was defined as the percentage of time in swing phase when one joint of a joint pair was moving while the other was not (Earhart and Bastian 2001). A joint was considered to be moving whenever its angular velocity reached or exceeded 5°/s. Thus for every frame of data collected, if the angular velocity of one joint of the joint pair was 5°/s or more, while if the angular velocity of the other joint was less than 5°/s, decomposition was said to be taking place. The total extent of decomposition was calculated by summing the number of frames where the joints were decomposed, dividing by the total number of frames in swing phase, and multiplying by 100. This indicated the total percentage of time in swing phase when decomposition occurred. Decomposition indices were calculated for the ankle-knee, ankle-hip, and knee-hip joint pairs. Finally, we calculated walking speed as stride length divided by the time it took to complete the stride.

We used the measures of balance (normalized weight-shifting distances) and visually guided leg-placement (foot endpoint errors) to categorize the cerebellar subjects. This was done by generating a 99% confidence interval from control group data for each task. We then identified whether the performance by each cerebellar subject fell within or outside of the confidence interval for each task. Cerebellar subjects whose performance fell outside the confidence interval for a task were said to have that deficit. This resulted in four separate subgroups within the cerebellar group: cerebellar subjects with no detectable deficits, a balance deficit only, a leg-placement deficit only, or both balance and leg-placement deficits. Confidence intervals are typically set at 95 or 99%; we chose the more conservative confidence interval so that there was only a 1% chance that "normal" performance would be outside this range (Fisher and van Belle 1993).

Three statistical tests were done. First, we compared performance between control and cerebellar groups during the balance and leg-placement tasks using Student's t-test for independent samples. Second, we compared walking performance among the control and cerebellar subgroups (e.g., balance deficit and leg-placement deficit subgroups) using the Kruskal-Wallis ANOVA by ranks test. We used this nonparametric test because of the relatively small sample sizes in each of the cerebellar subgroups. When the Kruskal-Wallis test yielded a significant effect, post hoc analysis was done using the Mann-Whitney U test, with a Bonferroni correction for multiple comparisons (Hayes 1994). Finally, we determined the extent to which walking speed could be predicted by our measures of balance and leg-placement impairments using a stepwise forward multiple regression analysis. Statistica (StatSoft, Tulsa, OK) and CoStat (CoHort Software, Berkeley, CA) software were used for all statistical analyses.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Balance and leg-placement tasks

As a group, cerebellar subjects demonstrated a decreased ability to voluntarily shift their weight laterally during the dynamic balance task. Mean normalized weight-shifting distances were 94.39 ± 5.15% (mean ± SE) for the control group versus 72.84 ± 6.96% for the cerebellar group (P = 0.0174). Likewise, cerebellar subjects demonstrated decreased accuracy in foot placement during the visually guided leg-placement task. Mean resultant foot endpoint errors were 16.03 ± 1.26 mm for the control group versus 27.97 ± 4.13 mm for the cerebellar group (P = 0.0088). Thus as a whole, the cerebellar group showed deficits of both balance and leg-placement capabilities.

Nevertheless, some subjects within the cerebellar group performed better than others during one or both of the tasks. Figure 2 shows the performance of two control and four cerebellar subjects during the balance or leg-placement tasks. The paths of the center of pressure from a control and two cerebellar subjects performing the balance task are depicted in Fig. 2A. Recall that the weight-shifting distances were normalized to foot spread measured at midfoot. Therefore weight-shifting distances of >100% were possible. Cerebellar subject CBL-18 showed little evidence of a balance deficit; he was able to shift his weight laterally between feet as well as the control subject. CBL-11, however, had significant balance impairments; he showed excessive sway in the anterior-posterior direction and did not shift his weight between feet as far as the control subject. Figure 2B shows the foot-path traces and foot endpoint errors for a control and two cerebellar subjects performing the leg-placement task. Cerebellar subject CBL-7 showed little evidence of dysmetria during this visually guided placement task. Her performance resembled that of the control subject with a relatively smooth and low step and minimal endpoint error. CBL-12, however, performed poorly. She took high, irregular steps and frequently overshot the target. Within the cerebellar group, we found that some subjects performed relatively well during both tasks, some performed relatively poorly during both tasks, and some showed a dissociation in performance between the two tasks, performing well during one but not the other.

View larger version (29K): [in this window] [in a new window]   Fig. 2. A: individual data from 3 subjects during the weight-shifting balance task. A bird's-eye view of the path of the center of pressure is shown for single trials for a control and 2 cerebellar subjects. The x axis represents movement of the center of pressure in the medial-lateral dimension. The y axis represents movement of the center of pressure in the anterior-posterior dimension. Graphs are scaled in millimeters. Average normalized weight-shifting distances are provided for each subject in the bottom right corner. B: individual data from 3 different subjects during the leg-placement task. Top: foot paths in the sagittal plane from start (left) to stop (right) on the target for a control and 2 cerebellar subjects. Three trials are overlaid. Bottom: a bird's-eye view of the endpoint locations of the foot marker relative to the target location for several trials. The open concentric circles represent radii of 10, 20, and 30 mm. The x axis represents error in the medial-lateral dimension. The y axis represents error in the anterior-posterior dimension. Graphs are scaled in millimeters. Average foot endpoint errors are provided for each subject in the bottom right corner.

Categorization

The categorization (balance vs. leg-placement deficits) for each of the cerebellar subjects is indicated in Table 1. The 99% confidence interval for the control group produced cutoff criteria of 79.65% for the normalized weight-shifting distance (measure of a balance deficit) and 19.63 mm for the resultant foot endpoint error (measure of a leg-placement deficit). Therefore cerebellar subjects who had weight-shifting distances <79.65% were considered to have a balance deficit, and those who had foot endpoint errors >19.63 mm were considered to have a leg-placement deficit. Of our sample of 20 subjects, 10 demonstrated an appreciable dissociation between balance and leg-placement deficits, with six subjects having a relatively isolated deficit of balance and four having a relatively isolated deficit of leg placement. Seven subjects showed deficits of both balance and leg placement. The remaining three subjects did not show any profound deficits of either balance or leg placement.

Control and cerebellar subgroup data for performance during the balance task, leg-placement task, and ICARS scores are shown in Fig. 3. As classified, subjects in the balance deficit subgroup performed poorly during the weight-shifting task (Fig. 3A) but as well as controls during the visually guided stepping task (Fig. 3B). Likewise, subjects in the leg-placement deficit subgroup performed poorly during the visually guided stepping task (Fig. 3B) but as well as controls during the weight-shifting task (Fig. 3A). Subjects in the neither deficit subgroup performed well during both tasks; subjects in the both deficits subgroup performed poorly during both tasks. ICARS scores (Fig. 3C) were not significantly different among the four cerebellar subgroups (P = 0.3211), probably because of the large within-group variability. Table 2 shows these same average ICARS scores broken down by subscore (posture and gait, limb kinetics, speech, and oculomotor deficits) for each of the cerebellar subgroups. Notice that subjects in the balance deficit only subgroup had higher scores (indicating worse performance) in the posture and gait component than subjects in the leg-placement deficit only subgroup. In contrast, subjects in the leg-placement deficit only subgroup had higher scores in the limb kinetics component than subjects in the balance deficit only subgroup. Thus the categorization, based on weight-shifting and leg-placement scores, appeared to match our clinical impression, based on ICARS subscores. However, when examining total ICARS scores, regardless of the specific type of impairment (Fig. 3C), subjects classified with neither deficit had relatively low scores; subjects classified with both deficits had relatively high scores, and subjects in the balance deficit and leg-placement deficit subgroups had similar ICARS scores, which were higher than those in the neither deficit subgroup but lower than those in the both deficits subgroup. The similar total ICARS scores between these two subgroups indicated that the overall level of severity of motor impairment was also similar. We therefore concluded that any differences in walking performance seen between these two cerebellar subgroups could be attributed to the specific type of motor deficit (balance vs. leg placement) rather than a difference in the overall severity of motor impairment. We restricted all subsequent comparisons of the specific features of walking to the cerebellar subjects in the balance and leg-placement deficit only subgroups compared with control subjects because we wanted to determine the specific walking abnormalities due to isolated deficits of either balance or voluntary leg control and because the neither deficit and both deficits cerebellar subgroups differed in their overall level of severity of motor impairment.

View larger version (21K): [in this window] [in a new window]   Fig. 3. Performance during the balance and leg-placement tasks and International Cooperative Ataxia Rating Scale (ICARS) scores for all groups. A: weight-shifting distances, normalized to foot spread, during the balance task. Cerebellar subjects in the balance deficit and both deficits subgroups had reduced lateral weight-shifting distances compared with the other groups. B: foot endpoint errors during the visually guided leg-placement task. Cerebellar subjects in the leg-placement deficit and both-deficits subgroups had greater errors than all other groups. C: ICARS scores for all cerebellar subgroups. Subjects in the leg-placement and balance deficit subgroups had similar ICARS scores, whereas subjects in the no deficits subgroup had slightly lower scores and subjects in the both deficits subgroup had slightly higher scores. However, these differences were not significant overall (P = 0.3211). Error bars represent ± 1 SE.

Walking

We next assessed whether cerebellar subjects with balance or leg-placement deficits had distinct walking abnormalities. Figure 4 shows an example of a control subject and two cerebellar subjects, one each from the balance deficit and leg-placement deficit subgroups. The figure shows each subject's performance on the balance task, leg-placement task, and joint angles during walking. The control subject's performance is typical for the group, demonstrating full excursion of the center of pressure during the balance task, minimal errors during the leg-placement task, and normal joint range of motion during walking. CBL-5 had a marked deficit in balance only; CBL-1 had a marked deficit in leg placement only. The joint angle traces for CBL-5 demonstrate several walking abnormalities: peak ankle plantarflexion, knee flexion, and hip flexion and extension are all decreased. CBL-5 also decomposes joint movements of the leg, particularly at the ankle and knee. The ankle joint normally plantarflexes and then dorsiflexes (indicated by open arrows), to accelerate the limb forward, and then clear the foot during swing. During this same period, the knee joint first moves into flexion, then extension (indicated by closed arrows). CBL-5 decomposes during this period: the ankle maintains a relatively constant angle while the knee continues to move. CBL-5 also walked slower (note the time scale changes) and was more variable in the timing and range of joint angles. In contrast, the amplitude, timing, and pattern of joint angles made by CBL-1 during walking closely resemble that of the control subject.

View larger version (34K): [in this window] [in a new window]   Fig. 4. A: performance during the balance and leg-placement tasks from a control subject (CNT-5), a cerebellar subject from the balance deficit subgroup (CBL-5), and a cerebellar subject from the leg-placement deficit subgroup (CBL-1). See Fig. 2 for axes labels. B: joint angles during walking for the same 3 subjects. Joint angles are shown for a full stride and are aligned on the time of heel strike. Several strides are overlaid for each subject. Positive angles represent flexion; negative angles represent extension. Arrows indicate the period of combined ankle (open arrows) and knee (filled arrows) movement, and illustrate ankle-knee decomposition by subject CBL-5.

Figure 5 shows stride lengths and stride widths for another control subject and two different cerebellar subjects, again one each from the balance deficit and leg-placement deficit subgroups. CBL-6 had an isolated deficit in balance; CBL-2 had an isolated deficit in leg placement. In the figure, the vertical line represents the average stride length for each subject; the horizontal line represents the average stride width. The symbol "R" indicates, the location of the ankle marker relative to the contralateral ("L") ankle marker on 5 consecutive walking trials. Stride lengths and widths were normalized to height. Stride lengths were decreased and more variable in CBL-6, the subject who had a balance deficit. Stride widths were also slightly increased in CBL-6 compared with the control subject. Performance by CBL-2, on the other hand, was similar to the control subject.

View larger version (12K): [in this window] [in a new window]   Fig. 5. Stride lengths and widths during walking from a control subject (CNT-2), a cerebellar subject from the balance deficit subgroup (CBL-6), and a cerebellar subject from the leg-placement deficit subgroup (CBL-2). Average stride lengths and widths are indicated by the vertical (stride length) and horizontal (stride width) lines. Individuals trials are indicated by the symbol "R" and are shown to illustrate the trial-to-trial variability in stride length and width. Each "R" represents the endpoint location of the ankle marker at heel strike, relative to the contralateral ankle ("L"), on subsequent trials. Average stride lengths (SL), stride widths (SW), and stride length variability (SL COV) are provided for each subject in the bottom right corner.

We next assessed the group data for all of the measured features of cerebellar gait ataxia. Recall that we predicted cerebellar subjects with a balance deficit would walk with a wide based, stiff-legged gait. For these subjects, we expected to see reduced stride lengths, increased stride widths, and reduced peak joint angles. We predicted cerebellar subjects with a leg-placement deficit would walk with a dysmetric, high-stepping gait. For these subjects, we expected to see increased stride-length variability, increased foot-path ratios, and increased joint-joint decomposition indices.

Figure 6 shows group data for the walking measures that we thought would be related to a balance deficit. Average stride lengths are depicted in Fig. 6A. Stride lengths were slightly reduced in the cerebellar leg-placement deficit subgroup, and markedly reduced in the cerebellar balance deficit subgroup (Kruskal-Wallis ANOVA by ranks, P = 0.0005). Post hoc Mann-Whitney U tests (corrected for multiple comparisons) revealed that the only significant difference was between the control group and the balance deficit subgroup (P = 0.0006). Stride widths are shown in Fig. 6B. Stride widths were approximately equal in the control group and cerebellar leg-placement deficit subgroup but slightly increased in the cerebellar balance deficit subgroup. However, these differences were not significant. Peak joint angles are shown in Fig. 6, C-E. At the ankle (Fig. 6C), peak plantarflexion was greatest in the control group, somewhat reduced in the cerebellar leg-placement deficit subgroup, and markedly reduced in the cerebellar balance deficit subgroup (P = 0.0170). At the knee (Fig. 6D), peak flexion was approximately equal in the control group and cerebellar leg-placement deficit subgroup, and reduced in the cerebellar balance deficit subgroup (P = 0.0072). At the hip (Fig. 6E), peak flexion was greatest in the control group, slightly reduced in the cerebellar leg-placement deficit subgroup, and markedly reduced in the cerebellar balance deficit subgroup (P = 0.0336). Post hoc analyses indicated that for each of the three joint angles, the only significant difference was between the control group and the balance deficit subgroup (ankle plantarflexion: P = 0.0042; knee flexion: P = 0.0029; hip flexion: P = 0.0131).

View larger version (16K): [in this window] [in a new window]   Fig. 6. Walking variables predicted to be most impaired by a balance deficit. Group means from control subjects (n = 20), cerebellar subjects in the leg-placement deficit subgroup, and cerebellar subjects in the balance deficit subgroup. The cerebellar balance deficit subgroup performed abnormally on most measures. A: average stride lengths, normalized to height (P = 0.0005). B: average stride widths, normalized to height (ns). C: peak ankle plantarflexion angles (P = 0.0107). D: peak knee flexion angles (P = 0.0072). E: peak hip flexion angles (P = 0.0336). Error bars represent ±1 SE.

Figure 7 shows group data for the walking measures that we thought would be related to a leg-placement deficit. Coefficients of variation for stride length are depicted in Fig. 7A. We were surprised to find that coefficients of variation were approximately equal in the control group and cerebellar leg-placement deficit subgroup but increased in the cerebellar balance deficit subgroup (P = 0.0245). Post hoc analysis revealed that the only significant difference was between the control group and the balance deficit subgroup (P = 0.0091). Foot-path ratios are shown in Fig. 7B. Path ratios were approximately equal in the control group and cerebellar leg-placement deficit subgroup but increased in the cerebellar balance deficit subgroup. This difference was not significant, however. Joint-joint decomposition indices are shown in Fig. 7 C-E. At the ankle-knee and ankle-hip pairs (Fig. 7, C and D), decomposition indices were low and approximately equal in the control group and cerebellar leg-placement deficit subgroup but increased in the cerebellar balance deficit subgroup although not significantly. At the knee-hip pair (Fig. 7E), decomposition indices were lowest in the control group, slightly increased in the cerebellar balance deficit subgroup, and even more elevated in the cerebellar leg-placement deficit subgroup (P = 0.0420). Here, the post hoc analysis indicated that the only significant difference was between the control group and the leg-placement deficit subgroup (P = 0.0119).

View larger version (14K): [in this window] [in a new window]   Fig. 7. Walking variables predicted to be most impaired by a leg-placement deficit. Group means from control subjects (n = 20), cerebellar subjects in the leg-placement-deficit subgroup, and cerebellar subjects in the balance-deficit subgroup. Contrary to our prediction, the cerebellar leg-placement-deficit subgroup performed abnormally on only 1 measure, and the balance-deficit subgroup performed abnormally on most measures. A: stride-length variability (the coefficient of variability; P = 0.0245). B: foot-path ratios during swing (ns). C: decomposition at the ankle-knee joint pair (ns). D: decomposition at the ankle-hip joint pair (ns). E: decomposition at the knee-hip joint pair (P = 0.0420). Error bars represent ±1 SE.

To summarize the findings from Figs. 6 and 7, the Kruskal-Wallis ANOVA by ranks showed that there were significant differences among the three groups in 6 of the 10 measures of walking: stride length, peak ankle plantarflexion, peak knee flexion, peak hip flexion, stride-length variability, and the knee-hip decomposition index. Post hoc Mann-Whitney U tests (corrected for multiple comparisons) revealed that the cerebellar balance deficit subgroup was significantly different from the control group for the first five of the six measures. The cerebellar leg-placement deficit subgroup was significantly different from the control group for only one measure (knee-hip decomposition index). We did not find significant differences among the three groups for the other four measures of walking: stride width, foot-path ratios, ankle-knee decomposition index, and ankle-hip decomposition index. Nevertheless, for each of these measures, the cerebellar balance deficit subgroup performed worse than the cerebellar leg-placement deficit subgroup.

Relationship between balance and leg-placement deficits and walking speed

Next, we assessed the extent to which balance and leg-placement deficits contributed to the overall walking deficit. We used speed as our measure of the overall walking deficit because decreased walking speed is considered a rather global indicator of walking impairment and is common to a number of different neurological conditions (Ebersbach et al. 1999; Levangie and Norkin 2001; Shumway-Cook and Woollacott 1995). Therefore we predicted that walking speed would be reduced in cerebellar subjects with either a balance or a leg-placement deficit. Using data from all of the cerebellar subjects, we performed a multiple regression analysis where walking speed was predicted based on our measures of balance (weight-shifting distances) and leg-placement (foot endpoint errors) deficits. We performed a stepwise forward regression using these two variables. The analysis produced a significant multiple R of 0.760 (P = 0.0006). The standardized partial regression coefficient () was significant and substantially greater for the balance measure (weight-shifting distance:  = 0.7011, P = 0.0004) compared with the leg-placement measure (foot endpoint error:  = 0.2998, P = 0.0741). This indicates that balance impairment explained more of the variance in and was a much better predictor of walking speed than leg-placement impairment. Figure 8 shows the relationship between observed walking speeds and those predicted by the regression analysis.

View larger version (16K): [in this window] [in a new window]   Fig. 8. Observed vs. predicted walking speed for all cerebellar subjects (n = 20). The multiple R of 0.760 had a significance level of P = 0.0006. Only the balance measure explained a significant proportion of the variance in walking speed (weight-shifting distance:  =0.7011, P = 0.0004). Therefore we concluded that performance on the balance task was a better indicator of walking speed than performance on the leg-placement task. The diagonal line indicates the fit of observed vs. predicted walking speeds produced by the multiple regression equation.

Cerebellar both and neither deficit subgroups

As mentioned previously, the cerebellar subjects in the neither deficit and both deficits subgroups were not included in the group comparisons of particular walking parameters (e.g., Fig. 6 and 7) because we were more interested in determining the specific walking abnormalities due to isolated deficits of either balance or voluntary leg control and because these subgroups differed in their overall level of severity of motor impairment (as measured by ICARS score). As one might expect, however, we found that the cerebellar neither deficit subgroup performed most like the control group and the cerebellar both deficits subgroup performed the worst. For example, average normalized stride lengths were 92.82% for the control group, 90.78% for the cerebellar neither deficit subgroup, and 61.09% for the cerebellar both deficits subgroup. Stride length coefficients of variation were 2.92% for the control group, 2.20% for the cerebellar neither deficit subgroup, and 15.78% for the cerebellar both deficits subgroup.

We also assessed whether, in the both deficits subgroup, walking speed was better correlated with balance or leg-placement deficits. Performance on the balance task was strongly correlated with walking speed (r = 0.865, P = 0.026), whereas performance on the leg-placement task was not correlated with walking speed (r = 0.061, P = 0.908); six of seven subjects in the both deficits subgroup were used in this analysis. The remaining subject (CBL-11) was somewhat of an outlier in the both deficits subgroup; he showed a moderate impairment of balance with a very severe impairment of leg placement and walked extremely slowly. We speculate his walking deficit was due more to the interaction between a moderate balance deficit and the very severe leg-placement problem. In his case, both features were important predictors of his walking speed.

Overall, the data from the cerebellar neither deficit and both deficits subgroups were in keeping with expectations (e.g., the cerebellar both deficits subgroup tended to walk poorly and the cerebellar neither deficit subgroup tended to walk well). The findings from the cerebellar both deficits subgroup were also consistent with that found in the two relatively isolated deficit subgroups (e.g., for the majority of subjects in the cerebellar both deficits subgroup, balance deficits were correlated with decreased walking speed but leg-placement deficits were not).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study is the first to describe the relative contributions of balance and visually guided leg-placement deficits to cerebellar gait ataxia in humans. When we compared walking abnormalities among control subjects, cerebellar subjects with balance deficits, and cerebellar subjects with leg-placement deficits, we found two interesting results: compared with the control group, cerebellar subjects with balance deficits were significantly impaired on 5 of 10 measures of walking, whereas cerebellar subjects with leg-placement deficits were significantly impaired on only 1 measure, and cerebellar subjects with balance deficits performed worse than cerebellar subjects with leg-placement deficits on 9 of the 10 measures of walking. Additionally, balance impairment, but not leg-placement impairment, explained a significant proportion of the variance in walking speed among all of the cerebellar subjects. We conclude that deficits of balance are a stronger indicator of cerebellar ataxia during uninterrupted level-ground walking than deficits of visually guided leg placement.

We had predicted that certain features of cerebellar gait ataxia would relate to balance impairments, whereas others would relate to leg-placement impairments. The measures of stride length and the peak angles at the ankle, knee, and hip joints were, as expected, abnormal in the cerebellar balance deficit subgroup. The other measure that we had anticipated would be abnormal in the balance-deficit subgroup was stride width. We did not see any significant differences in this measure among the groups, although our clinical impression was that the cerebellar subjects in the balance deficit subgroup did appear to stand with a wider base. The balance deficit subgroup did have the largest average stride width, but it differed from the control group by only approximately 10%. We speculate that the trial-to-trial variability in foot placement as well as the variability between subjects was too great to detect any differences with this sample size. Other studies have also failed to show a significant difference in stride widths among subjects with cerebellar disorders (Palliyath et al. 1998) or vestibulopathies (Krebs et al. 2002).

We incorrectly predicted that stride length variability, foot-path ratios and decomposition at the ankle-knee, ankle-hip, and knee-hip joint pairs would be abnormal in the cerebellar leg-placement deficit subgroup. Three of those measures, stride length variability, the ankle-knee decomposition index and the ankle-hip decomposition index, were, in fact, abnormal in the balance deficit subgroup and relatively unimpaired in the leg-placement deficit subgroup. Stride length variability was the only measure where there were significant differences between controls and the balance deficit subgroup. This is interesting because it is very similar to the measure we used to categorize leg-placement deficits; both stride length variability and foot-placement errors can be considered to reflect accuracy and consistency in limb placement, and the required joint kinematics and leg trajectories of the two tasks are nearly identical. Nevertheless, the same cerebellar subjects who had foot placement errors during the stepping task did not demonstrate inconsistency in foot placement during walking, whereas the cerebellar subjects who did not demonstrate errors during the stepping task (but had significant balance impairments) did produce irregular and inconsistent stride lengths. We conclude that control of leg movement, therefore, must have task-specific components to it. Presumably, locomotion engages different regions of the cerebellum than those engaged during a voluntary limb movement task. The reduced necessity for visual guidance during walking surely also contributed to differences between the two tasks. Although decomposition at the ankle-knee and ankle-hip joint pairs was not significantly different, we think it worth noting that they were more abnormal in the cerebellar balance deficit subgroup than the cerebellar leg-placement deficit subgroup compared with controls. We hypothesize that subjects in the balance deficit subgroup used decomposition as a strategy to decrease postural disturbances during walking. Stiffening the ankle joint would reduce push off forces, which act to propel the center of mass forward and can be destabilizing (for an example of this strategy, see subject CBL-5 in Fig. 4B).

The only measure where we saw significant abnormalities in the cerebellar leg-placement deficit subgroup was the knee-hip decomposition index. This is the only indication we found that cerebellar subjects with deficits of voluntary limb control may exhibit some abnormalities during walking that subjects with balance deficits do not. Decomposition is typically considered a strategy for reducing degrees of freedom during complex, multi-jointed movements, and this may explain why cerebellar subjects with a leg-coordination deficit decomposed at the hip and knee joints. However, it is curious that these same subjects did not need to utilize this strategy at the ankle-knee and ankle-hip joint pairs.

We also should point out that only four subjects fell into the leg-placement deficit subgroup (2 fewer than the balance deficit subgroup). Therefore it is possible that the smaller sample size in the leg-placement deficit subgroup prevented us from detecting more subtle differences between this subgroup and the control group. We cannot conclusively rule this out. We do, however, think that this number of subjects probably represents a realistic estimate of the incidence of cerebellar subjects with only voluntary leg-coordination deficits (i.e., 20% of cerebellar subjects). In our experience, the incidence of this type of deficit in isolation is fairly uncommon; most people with cerebellar damage have either impaired balance or both types of deficits.

Cerebellar contributions to human locomotion

Much of the normal locomotor pattern is likely generated by brain stem and spinal structures. However, the cerebellum may be responsible for constantly adjusting the precise intra- and interlimb coordination patterns required during locomotion and modulating various reflex patterns in accord with changes in the environmental context (Ito 1984). The anatomy of the cerebellum supports this hypothesis. In the cat, spinocerebellar tracts and spinoreticulocerebellar pathways project densely to the intermediate and medial cerebellum and carry signals related to rhythmical hindlimb movements. In turn, outputs from these same regions of the cerebellum project back to spinal cord pattern generators via the vestibulospinal, reticulospinal, and rubrospinal tracts (Arshavsky et al. 1983). These findings suggest an important role for the cerebellum in the control of locomotion and underscore how, anatomically, the more medial regions of the cerebellum seem particularly suited for this task. Indeed, other studies have shown that cerebellar walking abnormalities and impaired balance tend to be coupled, and that these types of movement impairments are related to damage to the medial zone of the cerebellum (Arshavsky et al. 1983; Botterell and Fulton 1938a,b; Chambers and Sprague 1955a,b; Orlovskii 1972; Thach et al. 1992; Yu and Eidelberg 1983).

This finding has been difficult to reproduce in humans. A few studies have characterized the specific patterns of postural sway in subjects with different types of focal cerebellar lesions (Diener et al. 1984; Mauritz et al. 1979). Clinical reports suggest that patients with lateral cerebellar damage tend to have prominent arm or leg ataxia, whereas patients with medial cerebellar damage tend to have more profound gait ataxia and impaired balance (Dichgans and Diener 1985). Our study is consistent with these reports and demonstrates that, like in animals, gait ataxia in humans is also tightly linked to impaired balance. However, because focal lesions were not common in our sample, we cannot make any strong claims as to which zone(s) were involved in these motor impairments. Nevertheless, the data do seem to support the animal literature. Six of the 20 cerebellar subjects had focal lesions where the damage was clearly more extensive in one of the cerebellar zones compared with the others (subjects CBL-5, -9, -10, -18, -19, and -20). Of those six subjects, all had the most significant damage to the medial cerebellar zone (by vermal hemorrhage or surgical vermal split). Three of those subjects had a balance deficit and fell into the balance deficit subgroup. None had evidence of a leg-placement deficit. Thus our data are consistent with the theory that the medial cerebellum is heavily involved in the control of balance and locomotion and less so in the control of visually guided movements of the distal limbs.

The remaining three of the six subjects who had focal medial cerebellar damage performed very well on both the balance and leg-placement tasks and fell into the neither deficit subgroup. These subjects all had the same diagnosis of medulloblastoma and had undergone the same surgical procedure, bisection of the posterior vermis. Not only were these subjects unimpaired in the balance and leg-placement tasks, their walking was virtually indistinguishable from control subjects. We think that there are a couple of possible explanations for why these three subjects did not show balance or walking abnormalities similar to the other three. First, the subjects with no deficits were all relatively young and tested at least 2 years after surgery. Thus they may have had a greater capacity to recover, possibly by increased dependence on other brain regions. This explanation is not completely satisfactory, however, because subject CBL-9 in the balance deficit subgroup was also young and tested 2 years after surgery. A second possibility is that lesion extent and location might account for the differences. Subjects in the cerebellar balance deficit subgroup who had focal vermal damage had either a hemorrhage (subjects CBL-5 and -10) or a split involving the superior portion of the vermis (CBL-9). Subjects in the cerebellar neither deficit subgroup all had splits involving only the posterior vermis. Another study examining five children who underwent surgical split of the posterior vermis also found few locomotor abnormalities during regular walking but more noticeable deficits when walking in tandem (Bastian et al. 1998).

Our data suggest that balance impairments contribute to cerebellar gait ataxia during uninterrupted walking over level ground more than leg-placement deficits. However, we predict that a voluntary leg-coordination deficit might affect walking in situations where stronger visual guidance is necessary. For instance, walking while stepping onto targets (e.g., stepping on rocks to cross a stream) or over obstacles (e.g., a fallen log) are instances where strong visual guidance for accurate leg trajectory and foot placement are required, in much the same way as was necessary to complete our leg-placement task. We predict that successful walking in these conditions would depend on both balance and leg-placement capabilities. The lateral region of the cerebellum has heavy projections to visual, parietal, premotor, and prefrontal areas of cortex (Stein and Glickstein 1992) and is known to help coordinate voluntary movements of limb segments (Gilman et al. 1976; Goodkin et al. 1993; Thach et al. 1992). Thus the lateral cerebellum is aptly suited for assisting with modes of walking where stronger visual guidance and/or novel adaptations to the locomotor pattern are required. The animal literature appears to validate this hypothesis. Neurons in the cat lateral cerebellum have been shown to fire much more so during a visually guided walking task (walking on a horizontal circular treadmill) or when visual guidance provides a behaviorally relevant cue for subsequent stepping patterns (when a rung on the horizontal ladder is elevated just prior to the cat's approach) as compared with uninterrupted walking over a level surface (Armstrong and Marple-Horvat 1996; Armstrong et al. 1997; Marple-Horvat et al. 1998). Schwartz and colleagues (1987) found that neurons in the cat dentate were relatively unresponsive and poorly modulated during unperturbed treadmill locomotion but more responsive and highly modulated in response to perturbations during locomotion, suggesting that the lateral cerebellum may also be particularly involved with monitoring alterations of the locomotor cycle.

In summary, we have shown that cerebellar subjects with impaired balance demonstrate most of the features classically described as gait ataxia, whereas cerebellar subjects with impaired visually guided leg-placement abilities do not. We also showed that performance on our balance measure predicts walking speed in subjects with cerebellar damage. Our findings are consistent with animal studies that suggest localization of the control of balance and uninterrupted level walking to the medial zone of the cerebellum (Chambers and Sprague 1955a,b; Thach et al. 1992). We postulate that different modes of walking, such as stepping onto visual targets (Armstrong and Marple-Horvat 1996; Crowdy et al. 2000; Marple-Horvat and Criado 1999) or navigating obstacles, would produce stronger associations between locomotor deficits and visually guided leg-placement impairments.