Feb 17, 1994 - Charlotte L. Shupert. Christine Jones- ... F. B. Horak (l~) - C. L. Shupert . C. Jones-Rycewicz .... (Thomas and Brown 1987). We hypothesized ...
Exp Brain Res (1994) 101:159-164
9 Springer-Verlag1994
J. Timothy Inglis 9 Fay B. Horak Charlotte L. Shupert. Christine Jones-Rycewicz
The importance of somatosensory information in triggering and scaling automatic postural responses in humans
Received: 17 February 1994 / Accepted: 13 June 1994 To clarify the role of somatosensory information from the lower limbs of humans in triggering and scaling the magnitude of automatic postural responses, patients with diabetic peripheral neuropathy and agematched normal controls were exposed to posterior horizontal translations of their support surface. Translation velocity and amplitude were varied to test the patients' ability to scale their postural responses to the magnitude of the translation. Postural response timing was quantified by measuring the onset latencies of three shank, thigh, and trunk muscles and response magnitude was quantified by measuring torque at the support surface. Neuropathy patients showed the same distalto-proximal muscle activation pattern as normal subjects, but the electromyogram (EMG) onsets in patients were delayed by 20-30 ms at all segments, suggesting an important role for somatosensory information from the lower limb in triggering centrally organized postural synergies. Patients showed an impaired ability to scale torque magnitude to both the velocity and amplitude of surface translations, suggesting that somatosensory information from the legs may be utilized for both direct sensory feedback and use of prior experience in scaling the magnitude of automatic postural responses. Abstract
Key words Posture 9 Somatosensory - Neuropathy E M G - Human
Introduction Somatosensory, vestibular, and visual sensory signals resulting from unexpected perturbations in stance are J. T. Inglis
Department of Physical Therapy, Elborn College, University of Western Ontario, London, Ontario, Canada N6G 1H1 F. B. Horak (l~) - C. L. Shupert . C. Jones-Rycewicz R.S. Dow Neurological SciencesInstitute, Good Samaritan Hospital, 1120 NW 20th Ave., Portland, OR 97209, USA; FAX no: (503) 229-7229
integrated by the central nervous system and trigger central programs that produce fast, appropriately scaled postural responses that restore equilibrium (Macpherson et al. 1986; Dietz 1992). The relative importance of the individual sensory systems in posture control is not yet fully understood and is likely to change depending on the task and context (Horak et al. 1989; Maki and Whitelaw 1993; Horak and Diener 1994). Nevertheless, the somatosensory system provides two types of sensory information which could contribute significantly both to triggering automatic postural responses and to scaling the magnitude of postural responses to the velocity and amplitude of postural disturbances. First, muscle proprioceptors and joint afferents signal joint position and movement, and second, mechanoreceptors in the sole of the foot signal the changing pattern of pressure and shear forces resulting from body movement. Both of these types of information could signal the onset and magnitude of any disturbance in equilibrium. Stimulation of somatosensory afferents in the lower limb by muscle vibration is known to result in postural adjustments (Enbom et al. 1988; Roll et al. 1989). Also, loss of somatosensory information from the lower limbs resulting from disease or experimental manipulation is known to result in postural control abnormalities (Magnusson et al. 1990a,b; Horak et al. 1990; Diener et al. 1984a). However, it is not known whether the abnormalities result from faulty triggering or faulty scaling of postural responses. The results of previous studies show that loss of somatosensory information from the feet and lower limbs leads to increased body sway. Elimination of mechanoreception in the feet by exposure to hypothermia results in increased body sway in response to calf muscle vibration and galvanic stimulation of the vestibular system (Magnusson et al. 1990a,b). Disruption of somatosensory information from the feet only, by ischemic block at the ankle, results in normal muscle response latencies, but increased hip sway in response to support surface translations (Horak et al. 1990). However, when the ischemic block is applied at the level of the
160 Table 1 Summary of results of clinical tests (A absent)
Subject
D.E. A.F. C.C. B.B. A.M. M.H. R.H. T.S. A.P.
Severity rank
1 2 3 4 5 6 7 8 9
Age (years)
51 63 63 57 57 59 67 48 56
Duration of diabetes
20 10 23 22 20 38 15 30 44
Vibration threshold (~tm)"
5.6 49.1 114.1 147.1 39.7 51.2 10.8 40.9 152.7
Sensoryb,(m/s)
Motor ~ (m/s)
R
L
R
L
34.1 A A A A A A A A
36.5 31.8 A A A A A A A
A 35.6 35.6 29.0 37.6 A A A A
30.9 36.2 28.6 30.5 A 26.6 A A A
a Normal values vibration threshold: 0.72_+0.08 gm b Normal values sensory (sural) nerve conduction: 44.0_+3.9 m/s c Normal values motor (peroneal) nerve conduction: 48.3 +4.7 m/s
thigh, which also blocks proprioceptive information of muscle proprioceptive information, as well as cutafrom the muscles of the lower leg, body sway increases neous and joint somatosensory information, from the dramatically (Diener et al. 1984a). These findings, cou- 9feet and lower legs to the central nervous system pled with the fact that vestibular-absent patients tested (Thomas and Brown 1987). We hypothesized that reducwith eyes closed show normal amounts of sway in quiet tion in somatosensory information would lead to destance, and normal response latencies and normal scal- layed postural responses and to impaired magnitude ing of response magnitudes to support surface transla- scaling. Preliminary results have been reported (Inglis et tions, suggest that somatosensory information is both al. 1993). necessary and sufficient to trigger postural responses with normal onset latencies (Horak et al. 1990; Allum et al. 1994). If this is the case, the increased sway associated Materials and methods with experimental disruption of proprioception in lower experimental protocols described here were approved by the leg muscles could be due to delayed triggering or im- The Internal Review Board of the Legacy Good Samaritan Hospital paired scaling of postural control responses. and Medical Center and were performed in accordance with the Previous research in normal healthy subjects (Diener 1964 Helsinki Declaration. All subjects were volunteers who gave et al. 1988; Horak et al. 1989) has shown that the magni- their informed consent prior to participating in the studies. Subtude of postural responses is scaled to the magnitude of jects were nine patients (six men, three women, ages 48-67 years) with symmetrical lower extremity, sensory peripheral neuropathy support surface translations by two different mecha- secondary to diabetes (see Table 1), and eight healthy, agenisms. Peripheral sensory information that encodes matched control subjects. All peripheral neuropathy patients were stimulus velocity is sufficient for velocity scaling; sub- recruited from the Neurology Service and Diabetic Clinic of the jects scale their responses to translation velocity equally Good Samaritan Hospital. Prior to their participation in the current study, patients were well when stimuli are presented in random sequences examined by a neurologist to assess their general neurological and in blocks of similar trials. Amplitude scaling, how- status and their functional mobility. Subjects with evidence of ever, requires prior experience with the expected stimu- visual or vestibular impairment were eliminated, and subjects lus amplitude. Subjects scale their responses to ampli- with significant medical, orthopedic, muscular, psychological, or neurological complications were excluded. Table 1 summatude poorly when different amplitudes are presented in other rizes the results of clinical testing. Sensory neuropathy was conrandom orders, but scale appropriately when blocks of firmed by bilateral measures of lower extremity sensory (sural) similar trials are presented. The work presented here and motor (peroneal) nerve conduction. Vibration sensitivity addresses two questions: (1) Are somatosensory affer- (threshold) of the great toe using a Vibratron II and a two-alternaforced-choice paradigm showed significantly reduced vibraents from the lower limb critical for triggering the auto- tive, tion sensitivity in all patients. All patients demonstrated mild dismatic postural responses following horizontal transla- tal muscular weakness but were able to stand on toes and heels tion of the support surface? (2) Are these afferents essen- and were independent in ambulation. Clinical examination tial for scaling the magnitude of postural reactions to showed that joint position sense was lost in the great toe in seven translations of varying velocity and amplitude ? In these of the nine patients, was reduced in one patient (A.F.), and was intact in one patient (D.E.). studies, patients with somatosensory loss secondary to Because there was considerable variability in the severity of diabetic peripheral neuropathy, but with intact vestibu- the neuropathy in the individual patients, patients were ranked in lar and visual sensory information, were exposed to pos- order of severity based on the results of clinical testing, and the terior horizontal support surface translations of varying patients are listed in Table 1 in order severity from least (D.E.) to (A.P.). The ranking was based primarily on sensory and amplitude and velocity. Diabetes results in impaired cir- most motor nerve conduction times. D.E., who had measurable sensory culatory support of peripheral nerves and can therefore conduction in both legs, was ranked least severe, followed by A.F., result in peripheral neuropathy, which disrupts the relay who had measurable sensory conduction in the left leg only. C.C.
161 and B.B., who had no measurable sensory conduction in either leg, but measurable motor nerve conduction in both legs, were ranked above patients with measurable motor conduction in one leg only. Patients with higher conduction velocities were ranked above patients with lower conduction velocities. Patients R.H., T.S., and A.P., who had no measurable sensory or motor nerve conduction in either leg, were ranked in order of their sensitivity to vibration in the great toe. Based on this ranking, the patients could be divided into three groups. D.E. and A.F. were the least impaired; R.H., T.S., and A.P. were the most impaired, and B.B., C.C., A.M., and M.H. were moderately impaired. Apparatus and methods have been described in detail elsewhere and will be summarized here (Diener et al. 1988; Horak et al. 1989). Subjects stood with their eyes open on two computercontrolled hydraulic force platforms that were translated backward to cause forward body sway. Backward translations of the support surface at four different velocities (10 cm/s, 15 cm/s, 25 cm/s, 35 cm/s, constant amplitude of 6 cm) and five different amplitudes (1.2 cm, 3.6 cm, 6 cm, 9 cm, 12 cm, constant velocity of 15 cm/s) were presented in blocks of seven like trials. Trial duration was 1 s, including 100 ms of background quiet stance prior to the perturbation. The intertrial interval was varied by the experimenter (10-20 s), and perturbation onset was delayed until subjects regained their initial position. Muscle activation patterns and timing were determined using electromyography, and response magnitude was quantified using surface reactive forces. Electromyograms (EMG) were recorded from six representative ankle, knee, and lower trunk antagonist muscle pairs of the left leg with 2-cm surface electrodes spaced 2 4 cm apart (medial gastrocnemius, GAS; biceps femoris, HAM; paraspinalis at the level of the iliac crest, PAR). Amplified EMG signals were band-pass filtered (70-2000 Hz) and full-wave rectified, low-pass filtered (100 Hz), and sampled (500 Hz). Muscle activation latency was identified by inspection of single trials and defined as the earliest time after translation onset that evoked EMG activity exceeded baseline levels by 2 standard deviations. Because there was no effect of translation velocity or amplitude on the muscle onset latencies (ANOVA, P>0.05), the mean onset latency for each muscle for each of the subjects was computed by collapsing across all 63 trials (7 trials in each of nine conditions). Means and standard deviations for each muscle were then computed for each group. Surface reactive torque under each limb was used to determine the magnitude of the postural response. Torques were quantified by calculating the slopes of the linear regressions of the first 75 ms of active torque (initial rate of change of torque). Scaling of postural responses was quantified as the slope of the linear regression
between translation velocity or translation amplitude and the rate of change of the active torque, t-tests (alpha = 0.05) were used to compare the slopes and elevations (intercepts) of the regression lines for normals and patients.
Results Electromyographic onset latencies All patients and normal subjects were able to maintain balance without assistance on all trials. Both patients and normal subjects activated their dorsal muscles in a distal-to-proximal pattern in response to the forward sway induced by the backward platform translations (Fig. aA). However, Fig. 1A shows that the onset of the muscle responses was characteristically delayed at the ankle, thigh, and trunk in the patients. The onset of activity in GAS occurred at 116+11 ms in normals, compared with 133___9 ms in the patients (Fig. 1B). Delays were similar in HAM (154_+25 ms compared with 174+_27ms) and PAR (173+_28ms compared with 202_+ 22 ms). These differences were statistically significant (P < 0.05).
Fig. 1 A Typical averaged EMG recordings from one normal subject (top traces) and one representative neuropathy patient (bottom traces) from the three dorsal muscles activated by a backward platform translation of 35 cm/s, 6 cm in amplitude. Thefirst vertical line on the left represents the onset of platform movement. The three dotted vertical lines mark the onsets for each of the muscle groups for the normal subject. Note that for the neuropathy patient, the onset latencies for each muscle group are delayed. B Mean muscle onset latencies (milliseconds) and standard deviations for the dorsal muscles from normal (black columns) and neuropathy subjects (grey columns). Asterisk represents a significant difference at the P < 0.05 level. (GAS medial gastrocnemius, HAM biceps femoris, PAR paraspinalis, PLATplatform)
B
A
250
GAS
~i ~ _
.
~ ~ L 1 ~ ~ r ~ m~
1a
~l
lkTnrrn,~l
Subject Neuropathy Patient
~-, 200
HAM
150
PAR
i
1 38 laV
i ~
138 V
9
100
tO
1 50 I.tV PLAT
~
onset
translation ~ backward 6 cm at 35 cm/sec
~
4 Displacement
50
I
400
T i m e (ms)
800
GAS
HAM
PAR
Muscle Groups
162
B Amplitude Scaling
A Velocity Scaling 400 250 ~D
t~ o
300
o o ,---,
cD
~o . 200
xz ~ 200 r,.)
=z
o
o
,,
T
4.o%t--'~"
l
r,.)
150
100
0
I
I
I
I
I
I
I
I
I
I
5
10
15
20
25
30
35
40
2
4
Platform Velocity (crrds)
Fig. 2 Scaling of torque response to platform velocity (A) and amplitude (B) in normal subjects (circles) and neuropathy patients (squares). Linear regressions are based on the group mean values. Mean and standard error are shown for each value. Patients with neuropathy scale responses both to platform velocity and amplitude, but at a reduced level when compared with age-matched control subjects
Scaling torque responses
Velocity scaling The scaling of postural response magnitude to translations of varying velocity has been shown to be dependent upon sensory feedback (Diener et al. 1988) and to be uninfluenced by prior experience (Horak et al. 1989). If feedback from somatosensory afferents from the lower extremity is crucial for this velocity scaling, then patients with decreased somatosensation should have difficulty appropriately scaling response magnitude to translation velocity. Figure 2A shows the group mean and standard error of the initial rate of change of torque in response to four translation velocities for both the neuropathy (squares) and normal groups (circles). Both patients and normal subjects were able to scale their postural responses to platform velocity. The slope of the patients' regression was 4.8, and the normals' slope was 7.1. Both slopes were significantly different from zero (P < 0.001). However, the patients' slope was significantly lower than the normals' slope (P = 0.02), indicating decreased scaling in the patients. The elevations (intercepts) of the regression lines for normals and patients were not significantly different. The results of this experiment suggest that the patients' velocity scaling and force metrics may be correlated with the severity of their neuropathy. The three
6
I
I
f
I
8
10
12
14
Platform Amplitude (cm) subjects with the most severe impairments showed the lowest slopes and y-intercepts, while the two least impaired subjects showed the highest slopes and y-intercepts. The mean slope for the three most-impaired subjects was 2.9, compared with 7.7 for the least-impaired two subjects, and the y-intercept for the three most-impaired subjects was 74, compared with 207 for the leastimpaired subjects. Spearman's rank correlation was used to determine the relation between the severity of neuropathy and the slope and intercepts for the patients. For the slope, the correlation coefficient (rho) was -0.64, which is consistent with decreasing slope with increasing severity, and this coeffient approached statistical significance (P < 0.07). The correlation coefficient for the y-intercept was -0.73, consistent with decreasing intercept with increasing severity, and this correlation was statistically significant (P < 0.04).
Amplitude scaling Since the postural responses to all but the smallest translation amplitudes must be initiated prior to the completion of platform translation, subjects must predict translation amplitude based on previous translations; thus, the ability to scale responses to the amplitude of platform movement depends on prior experience with that amplitude (Horak et al. 1989). It is not known whether somatosensory information from the legs is critical for this prediction. Figure 2B shows the group mean and standard error for the initial rate of change of torque for each of the five translation amplitudes for both the neuropathy (squares) and normal groups (circles). Both neuropathy and normal subject groups were able to scale postural response magnitude to translation
163 amplitude (slopes significantly greater than zero, were highly reliable, they were not large enough to reP < 0.001). Similar to the results for velocity scaling, the sult in a total failure of postural responses; all the papatients' slope (1.9) was lower than the normals' (4.0), tients were able to maintain balance in response to the and this difference approached statistical significance surface translations tested in these experiments. Mag(P =0.06). The regression lines shown in Fig. 2B also nusson et al. (1990a) have estimated that adequate posshow that the patients' responses were generally larger tural control can be maintained (assuming a body than those of normals at this platform velocity, but height of 1.7 m and inverted pendulum sway) if feedback again, this difference only approached statistical signifi- delays do not exceed 300 ms, which is well beyond the range of delays found in the present study. cance (P = 0.06). Although neuropathy patients with reduced soThe lack of statistical significance in the differences between patients and normals for amplitude scaling matosensory inputs from the lower legs were still capaprobably resulted from the fact that the patients' perfor- ble of scaling the magnitude of their postural responses mance for amplitude scaling was fairly variable (see to both increasing velocity and amplitude of platform standard error bars in Fig. 2). However, the results of translation, their scaling was poorer than normals. The this experiment suggest that differences in disease sever- fact that the patients scaled their responses to varying ity may account for the variability. Spearman's rank velocity more poorly than normals is not surprising. correlations were performed for severity of neuropathy Somatosensory information from the legs could fully and amplitude slopes and y-intercepts for the patients. characterize the translation velocity early enough in the The correlation coefficient for severity of neuropathy perturbation to permit velocity scaling in normal suband amplitude slope was 0.67, which approached statis- jects. Even patients with mild diabetic neuropathy show tical significance (P < 0.06) and suggests increasing scal- abnormal neural coding of tactile stimuli (Mackel 1989), ing with increasing severity. This correlation is opposite and the patients studied here had pronounced soto that for severity of neuropathy and velocity scaling. matosensory losses. The patients' decreased amplitude The mean amplitude scaling slope for the two least-im- scaling is, however, somewhat surprising. In this experipaired subjects was -0.5, compared with 2.6 for the mental protocol, the platform movement in some condithree most impaired subjects. The correlation coefficient tions is not complete prior to the onset of automatic for the severity of neuropathy and y-intercept was postural responses, and hence magnitude scaling in -0.82, indicating decreasing y-intercept with increasing these conditions must rely on learning during repeated severity, and this correlation was statistically significant exposure to each amplitude, and not on somatosensory (P
