How do persons with chronic low back pain speed up and slow down?

Trunk–pelvis coordination and lumbar erector spinae

activity during gait§

Claudine J.C. Lamoth *, Andreas Daffertshofer, Onno G. Meijer, Peter J. Beek

Institute for Fundamental and Clinical Human Movement Sciences, Van der Boechorststraat 9,

Vrije Universiteit, 1081 BT Amsterdam, The Netherlands

Received 18 February 2005; accepted 24 February 2005

Abstract

In healthy walking, the timing between trunk and pelvic rotations, as well as erector spinae (ES) activity varies systematically with walking

velocity, whereas a comparable velocity-dependent adaptation of trunk–pelvis coordination is often reduced or absent in persons with low

back pain (LBP). Based on the hypothesis that trunk–pelvis coordination is linked to overall gait stability, persons with LBP can be expected to

have difficulties in dealing with perturbations. We examined the ability of 12 persons with LBP and 12 controls to adapt trunk and pelvis

rotations and ES activity to sudden changes in velocity. 3D angular movements of thoracic, lumbar, and pelvic segments and surface EMG

were recorded during treadmill walking at six different velocities, which increased or decreased unexpectedly. Relative phases of segmental

rotations were determined and (in-)variant properties of kinematics and ES activity were studied using principal component analysis.

Compared to healthy controls, persons with LBP exhibited a reduced ability to adapt trunk–pelvis coordination and ES muscle activity to

changes in velocity. Altered coordination and muscular control may reflect an attempt to stabilise the spine and prevent the occurrence of

unexpected perturbations. The assessment of gait patterns in terms of coordination may help clinicians to quantify movement impairments and

may suggest interventions aimed at facilitating the emergence of desired coordination patterns.

# 2005 Elsevier B.V. All rights reserved.

Keywords: Locomotion; Low back pain; Coordination; Variability; Principal component analysis

www.elsevier.com/locate/gaitpost

Gait & Posture 23 (2006) 230–239

1. Introduction

Many persons with chronic low back pain (LBP)

experience problems with walking. On average, they walk

slower than healthy walkers [1–3]. A slower preferred

walking velocity is a characteristic of many different gait

disorders [4–7], and generally hint at poor motor control.

Healthy walkers compensate for internal and external

perturbations that potentially disrupt walking. These flexible

reorganizations are possible because of the high degree of

coordination between cyclically moving body segments (e.g.

limbs, pelvis, trunk, and head) that characterises walking.

Walking appears to be composed of quite steady coordination

§ ESMAC 2004 Best Paper Award.

* Corresponding author. Tel.: +31 20 5988522; fax: +31 20 5988529.

E-mail address: C_Lamoth@fbw.vu.nl (Claudine J.C. Lamoth).

0966-6362/$ – see front matter # 2005 Elsevier B.V. All rights reserved.

doi:10.1016/j.gaitpost.2005.02.006

modes, specific phase, and frequency relations between

cyclical movements of limbs, pelvis, trunk, and head. In

unimpaired gait, these interactions or couplings are relatively

stable, yet adapt flexibly to changes in, for example, walking

velocity. The stability and flexibility of these coordination

patterns is reflected in their variability, as assessed through

various statistical procedures (e.g. co-)variance, coherence,

principal component analysis, PCA) [8–13].

Coordination between trunk and pelvis and the activity of

the associated musculature such as erector spinae (ES)

muscles have proven to be a useful entry point in the study of

human gait. When walking speed is varied, timing and

variability of trunk–pelvis coordination and ES activity

change systematically, presumably to cope with perturba-

tions and to preserve stable gait patterns [14–16]. In

unimpaired gait, increasing walking velocity changes the

phase difference, or relative phase (RP), between transverse

C.J.C. Lamoth et al. / Gait & Posture 23 (2006) 230–239 231

thoracic and pelvic rotations from more or less in-phase

(synchronous rotations in the same direction) toward more

antiphase coordination (synchronous rotation in opposite

direction) [11]. These segments also become more tightly

coordinated and less variable in the frontal plane. Similarly,

with increasing walking velocity the rather variable lumbar

erector spinae (LES) activity displays a biphasic activity

pattern with peak activity around foot contact and has little

activity during swing phases [17,18]. Compared to healthy

persons, persons with chronic LBP show a diminished

adaptation of trunk–pelvis coordination to increasing velocity

in that the relative phase between thoracic and pelvic rotations

remains more biased toward in-phase coordination. In those

persons, impaired trunk–pelvis coordination is associated with

poorly coordinated LES activity [17,19]. It is thus conceivable

that a slow preferred walking velocity of persons with LBP is

associated with an inability to adapt trunk–pelvis coordination

to velocity changes. In previous studies, velocity was

gradually increased to enable the subjects to accustom

themselves to higher velocity levels in a gentle and uniform

manner. Random changes in walking velocity disrupt the

coordination in a different and unexpected way each time and

can therefore be supposed to threaten overall gait stability

more than a gradual increase in velocity that has been used in

previous experiments. Persons with LBP are thus expected to

have more difficulties with those perturbations than healthy

persons. To test this expectation, the invariant and variant

properties of trunk–pelvis coordination and LES activity

patterns following perturbations in walking velocity were

studied and compared between groups.

2. Methods

2.1. Subjects

Twelve persons with chronic LBP (seven women, five

men) and 12 healthy subjects (five women, seven men)

participated. All subjects had participated in a previous

experiment in which velocity was gradually increased [17].

Only those LBP subjects who were able to walk at 6.2 km/h

participated in the present experiment. The Ethics Commit-

tee of the Medical Center of the Vrije Universiteit approved

the procedure and all subjects gave their written, informed

consent before participation. LBP subjects were assessed by

an orthopedic surgeon and were enrolled if they had (1) non-

specific LBP with pain and symptoms persisting for longer

than 3 months for which medical treatment had been sought;

(2) age between 18 and 65 years; and (3) ambulation without

a walking aid. Participants were excluded if they had (1)

LBP of traumatic or structural origin; (2) LBP with

neurological symptoms or pain radiation in the lower leg(s);

(3) previous back surgery; (4) spinal tumours or infections;

or (5) neurological and/or musculoskeletal disorders

unrelated to LBP. The healthy subjects had no history of

LBP or any other musculoskeletal disorders.

2.2. Procedure

Subjects walked for a few minutes on the treadmill at

different velocities to familiarise themselves and to

determine the subjects’ comfortable walking velocity.

Recordings were then performed at six velocities in a fixed

order: 6.2, 1.4, 3.8, 5.4, 2.2, and 4.6 km/h. Recordings

started immediately without allowing the subject to become

habituated to the new velocity and lasted 30 s for each

velocity. The participants could not predict whether the

velocity would increase or decrease.

Angular rotations of the trunk segments were recorded

using a 3D active marker movement registration system

(Optotrak 3020, Northern DigitalTM, Ontario, Canada).

Clusters of three markers were fixed on a light plate mounted

on rigid fixtures, which were designed to span the ES muscle

and spinous processes. These clusters were attached to the

trunk with neoprene bands at the level of the third thoracic

vertebra (T3), second lumbar vertebra (L2), and the sacrum.

Markers were also placed on the heels and the fifth

metatarsophalangeal joint to determine timing of events in

the gait cycle.

Before electrode placement the skin was shaved and

cleaned with alcohol. EMG activity from the LES muscle

left and right of L2 and the fourth lumbar (L4) processes was

recorded with pairs of surface electrodes (Blue sensor N-00-

S, Medicotest, Denmark: AG/AgCl discs, 1 cm diameter,

2 cm interelectrode distance) using a 16-channel portable

EMG recording device (Porti5, TMS-internationalTM,

Enschede, The Netherlands). Electrodes were placed at a

distance of 3 cm lateral from the vertebral column [20].

Kinematic data were sampled at 100 Hz and EMG data at

1 KHz. Both types of recordings were synchronized via

single trigger pulse. All raw data were exported for off-line

analyses in Matlab 6.50 (Mathworks, Natic, MA, USA).

We analyzed the kinematic data in a xyz-coordinate system

with x-axis along the line of progression, y-axis pointing

sideways and z-axis vertically upwards. Initially, a reference

measurement was conducted with the subject standing upright

providing a global reference system. From every marker

cluster a segment frame was defined and aligned with this

global reference system. In each segment frame the angular

rotations were calculated. Heel strikes were estimated via the

minimal vertical velocity of the toe marker [21].

EMG data were rectified using the Hilbert transform.

Kinematic and EMG data were low-pass filtered using a

fourth order bi-directional Butterworth filter with a cut-off

frequency of 10 and 20 Hz, respectively.

2.3. Data analyses

Time-dependent changes of the relative phases between

trunk segments were addressed through a Fourier transform

within a finite frame size e, which was shifted in time

[11]. Phases were obtained at the movement frequency v0

C.J.C. Lamoth et al. / Gait & Posture 23 (2006) 230–239232

of the proximal segment. We computed for a time series

x(t)

xðtÞ 7! xeðv; tÞ ¼Z 1

�1xðtÞWeðt � tÞe�ivðt�tÞ dt

with WeðtÞ ¼1 for 0 � t< e

0 otherwise

�(1)

(frame size e was fixed at twice the corresponding period

length: 2T0 = 4p/v0) and defined the phase as:

’eðv; tÞ ¼ arctanImfxeðv; tÞgRefxeðv; tÞg

� �(2)

Hence, the RP between the signals x(t) and y(t), representing

thoracic and pelvic (RPthpe) or lumbar and pelvic (RPlupe)

segmental oscillations in the transverse or frontal plane, was

given as:1 D’ðx;yÞv0

ðtÞ ¼ ’ðxÞ2T0

ðv0; tÞ � ’ðyÞ2T0

ðv0; tÞ. Finally,

for every RP we calculated the variance yielding a measure

for the variability of the coordination of interest.

Principal component analysis was applied to detect

similarities and deviations between LBP and control

subjects in adaptations to changing velocity [9]. Firstly,

PCA was applied to time-varying angles of thoracic, lumbar

and pelvic segments in the frontal and transverse plane and

to LES recordings over several gait cycles to identify

coherent patterns over subsequent stride cycles. To compare

LBP with control subjects we combined kinematic or LES

recordings into a single vector ~q ¼~qðtÞ, that is

~q ¼

~qcontrol1

..

.

~qcontrol12

~qLBP1

..

.

~qLBP12

0BBBBBBBBBBB@

1CCCCCCCCCCCA

; ~qsubject1;...;24¼~qs ¼

~qs;velocity1

~qs;velocity2

..

.

~qs;velocity6

0BBBBB@

1CCCCCA;

~qs;velocity1;...;6¼~qs;v ¼

qs;v;r¼1

qs;v;r¼2

..

.

qs;v;r¼6

0BBBB@

1CCCCA (3)

In words, the ~q contained components qs;v;r each repre-

senting the data of a certain time series r (r = 1, . . ., 6

referring to transverse and frontal plane thoracic, lumbar,

and pelvis rotations, and r = 1, . . ., 4 referring to left, right

L2 and L4 ES recordings) of a certain subject s (s = 1, . . .,24) walking at a velocity v (v = 1, . . ., 6). For~q (or the set

{qs;v;r}) we computed the covariance matrix and its eigen-

vectors ~vðkÞ and eigenvalues lk, referred to as principal

components (PCs). The PCs were sorted in descending

order of lk according to the amount of data variance

accounted for by each component—lk measures the var-

1 Note that a phase difference of 1808 describes antiphase and 08 in-phase

coordination.

iance along the direction~vðkÞ. The time series along every

direction or mode is given as projection: jkðtÞ ¼~vðkÞ~qðtÞ.The number of PCs representing the coherent signal struc-

tures was determined by visual inspection of the eigenva-

lue spectra using discontinuities in the spectra and the form

of the corresponding times series as cut-off criteria. The

contribution of a segment or EMG recording r of a LBP or

control subject s at a velocity v to the kth PC was given by

the coefficient vðkÞs;v;r that varied between �1 and 1. The

larger the absolute value of vðkÞs;v;r, the more the time series

qs;v;r contributed to the pattern k. In general, variations in

amplitude between variables will have a large effect on the

determination of PCs. In contrast, when data are normal-

ised by rescaling the individual signals to unit variance, all

variables have equal weight so that the determination of

PCs only depends on the underlying coordination pattern.

PCA was performed on normalised as well as non-normal-

ised kinematic data.

Secondly, to examine in detail the effect of LBP on

variability PCA was used as a data driven filter [9]. Using the

stride intervals, rescaled kinematic and LES recordings

obtained at every velocity were split into subsequent time

series containing a single stride cycle (six per velocity) each,

and thereafter time normalised to 0–100% of the stride cycle

using a cubic spline interpolation. We applied PCA to the

combined data of LBP and control subjects. The invariant

(global) pattern was defined as sum of the dominant PCs and

separated from the variant (residual) components in the set by

projecting the data onto all the remaining PCs. We wrote

~qðtÞ ¼~qðglobalÞðtÞ þ~qðresidualÞðtÞ and analyzed~qðglobalÞðtÞ and

~qðresidualÞðtÞ in terms of their variances. The variability of

residual and global patterns was examined separately for each

LBP and control subject, along time series of individual

segments and movement direction or muscles during stride

cycles obtained at one velocity. To relate the variability of LBP

subjects to the control group, the ratio between the mean

variability of each segment or LES recordings per velocity of

individual LBP subjects and that of the entire control group

was calculated [15]. This ratio provided an index of variability

for every LBP subject and segment or muscle per velocity.

2.4. Statistical analysis

Continuous RPs were averaged for every participant and

velocity. The circular variance of the RP was transformed

into a linear standard deviation (S.D.) that could be

subjected to conventional statistical tests [22]. Unpaired t-

tests were performed to examine differences in comfor-

table walking velocity, stride length, age, weight, and

length between LBP and control subjects. Statistical

differences between LBP and control subjects at walking

trials were assessed by computing ANOVAs for repeated

measures with six levels (walking velocity) as within-

subjects and HS (control versus LBP) as between-subjects

factor. The dependent variables were mean and S.D. of

RPs, absolute eigenvector coefficients of the PCs, and

C.J.C. Lamoth et al. / Gait & Posture 23 (2006) 230–239 233

Table 1

The effect of changing walking velocity (Vel) and HS (control vs. LBP) on the mean relative phases (RP, the standard deviation of the relative phases (S.D. RP),

and significant effects of the post-hoc analyses

Segment Repeated measures ANOVAs Post-hoc analyses HS

Vel Vel � HS HS 6.2 km/h 1.4 km/h 3.8 km/h 5.4 km/h 2.2 km/h 4.6 km/h

F5,110 P-value F5,110 P-value F1,22 P-value t(22) P-value t(22) P-value t(22) P-value t(22) P-value t(22) P-value t(22) P-value

Transverse plane

Mean RP

THPE 67.9 <0.01 2.5 0.03 10.6 <0.01 5.6 <0.01 2.2 0.04 4.5 <0.01

LUPE 25.5 <0.01 2.5 0.03 11.5 <0.01 4.3 <0.01 2.5 0.02 4.6 <0.01

S.D. RP

THPE 2.8 0.02 ns 7.4 0.01 2.6 0.02 2.3 0.03 2.1 0.0 1.9 0.07

LUPE 6.6 <0.01 ns 5.6 0.03 2.6 0.02 2.3 0.03 2.1 0.04

Frontal plane

Mean RFP

THPE ns ns ns

LUPE ns ns ns

S.D. RP

THPE 18.6 <0.01 4.8 <0.01 14.6 <0.01 �3.7 <0.01 �2.7 0.01 �2.4 0.03

LUPE 8.4 <0.01 2.1 0.06 8.7 <0.01 �2.4 0.03 �2.5 0.03 �2.5 0.03

variances of global and residual patterns of segment

rotations and LES activity. When significant main or

interaction effects were present, post-hoc t-tests were

performed (P < 0.05, two-sided was used as significance

level).

Fig. 1. Mean relative phase (RP) (left panels) and the S.D. of the RP (right panels)

pelvic rotations in the transverse (upper panels) and frontal plane (lower panels)

3. Results

On average, the LBP subjects were 36.8 years old

(S.D. = 10.9), weighted 72.4 kg (S.D. = 14.5), and were

1.74 m (S.D. = 0.11) tall. The controls were on average 30

as a measure of coordinative stability, between thoracic–pelvic and lumbar–

. Error bars indicate S.E.

C.J.C. Lamoth et al. / Gait & Posture 23 (2006) 230–239234

years old (S.D. = 8.1), weighed 73.3 kg (S.D. = 16.6) and

were 1.80 m (S.D. = 0.12) tall. Neither age, weight, nor

length was significantly different between groups. Comfor-

table walking velocity was significantly lower (t = 3.4,

P < 0.01) in the LBP (mean = 3.6 km/h, S.D. = 0.98) than in

the control group (mean = 4.8 km/h, S.D. = 0.68). Stride

length was significantly shorter for LBP subjects at 6.2 km/h

(t = 2.4, P = 0.02).

3.1. Relative timing between segments

Mean RPthpe and RPlupe in the transverse plane decreased

and increased significantly with velocity (Table 1 and

Fig. 1). This velocity effect was less pronounced for LBP

subjects, as was evidenced by a significant effect of HS and

HS � velocity interaction. At the three highest velocities the

change towards antiphase coordination of RPthpe was

reduced in the LBP group, while RPlupe remained more

or less in-phase at all velocities. The S.D. of RPthpe and

RPlupe was significantly smaller in the LBP than in the

control group, particularly a large velocity change was

followed by a lower variability in the LBP than in the control

Fig. 2. The effect of changing walking velocity on the variability of the residual a

residual or global pattern of each LBP subject divided by that of the control group.

that of the residual patterns. Error bars indicate S.E.

group. RPthpe and RPlupe in the frontal plane were not

significantly affected by velocity in both groups. Decreases

in velocity increased the variability of the RPs in the frontal

plane, whereas increases in velocity decreased variability.

With large velocity changes, the increase in the S.D. of

RPthpe and RPlupe was markedly larger for the LBP than for

the control group (Table 1 and Fig. 1).

3.2. Patterns of trunk coordination and variability

The first four PCs (global pattern) covered 85%

(51% + 29% + 3% + 2%) of the variance of the non-

amplitude-normalised data set and 78% of the amplitude-

normalised data set (43% + 28% + 4% + 3%). PC1 and PC2

represented trunk and pelvis oscillations at the stride

frequency and PC3 and PC4 oscillations at twice the stride

frequency. The corresponding coefficients revealed that the

LBP and control group contribution to the first four PCs did

not differ significantly. The variability of global patterns of

all segment rotations in both planes was not significantly

affected by HS; mean ratios were almost 1 at all velocities

(Fig. 2).

nd global pattern. The ratio over subjects was calculated as the S.D. of the

Open markers represent the variability of the global pattern, closed markers

C.J.C. Lamoth et al. / Gait & Posture 23 (2006) 230–239 235

Table 2

Significant effects of walking velocity (Vel), and HS (control vs. LBP group) on the variability of the residual pattern of transverse (trans.) and frontal (front.)

plane segment rotations, and significant effects of post-hoc analyses

Residual

pattern

Repeated ANOVA Post-hoc analyses HS

Vel Vel � HS HS 6.2 km/h 1.4 km/h 3.8 km/h 5.4 km/h 2.2 km/h 4.6 km/h

F(5,110) P-value F(5,110) P-value F(1,22) P-value t(22) P-value t(22) P-value t(22) P-value t(22) P-value t(22) P-value t(22) P-value

Thorax

trans.

7.1 <0.01 2.8 0.02 7.9 0.01 3.0 <0.01 2.9 <0.01 2.5 0.02 2.3 0.03

Lumbar

trans.

8.4 <0.01 2.2 0.03 16.2 <0.01 2.1 0.04 2.2 0.03 3.2 <0.01 2.2 0.04 3.3 <0.01

Pelvis

trans.

3.5 <0.01 9.7 <0.01 4.7 0.04 �2.3 0.03 �2.4 0.02 2.2 0.04 2.7 0.01 �2.6 0.02

Thorax

front.

3.5 <0.01 2.1 0.03 8.4 <0.01 �2.9 <0.01 �2.3 0.01 �2.5 0.01 �2.5 0.01

Lumbar

front.

3.1 <0.01 5.1 <0.01 7.7 0.01 �4.7 <0.01 2.1 0.04 �2.7 0.01 �2.3 0.01 �3.3 <0.01

Pelvis

front.

7.9 <0.01 6.3 <0.01 5.1 0.03 �2.7 0.01 �2.6 0.01 �2.5 0.01

Velocity manipulations significantly affected the varia-

bility of residual patterns of all segments in both planes of

motion. We did observe significant effects of HS and

HS � velocity for all segments (Table 2). The variability of

residual patterns of transverse thoracic and lumbar rotations

decreased with shifts to lower velocities and increased again

at the next higher velocity. An opposite pattern was found in

the frontal plane. Compared to the control group, residual

Fig. 3. Individual time series of superimposed stride cycles of left LES recorded at

6.2, 3.8, and 4.6 km/h. EMG recordings are amplitude normalised.

variability was smaller in the LBP group for transverse plane

rotations and larger for frontal plane rotations at all

velocities. Furthermore, the velocity-induced increase

(transverse plane) and decrease (frontal plane) in variability

was less pronounced in the LBP group (Table 2 and Fig. 2).

Residual variability of transverse pelvis rotations was

elevated at 3.8 and 2.2 km/h in the control subjects but

reduced in the LBP group, while in the frontal plane the

the level of L4 for one control and two LBP subjects at walking velocities of

C.J.C. Lamoth et al. / Gait & Posture 23 (2006) 230–239236

variability was significantly higher in the LBP than in the

control group at 1.4, 2.2, and 4.6 km/h.

3.3. Lumbar erector spinae activity patterns and variability

Initially LES activity appeared to be more variable in the

LBP group than in the control group due to a variety of

factors, including larger phase shifts, additional frequencies,

and prolonged activity around heel strike (Fig. 3).

We analyzed these variability patterns utilizing PCA. The

first six PCs described the global activity pattern covering

about 40% of the data’s variance. Further, we collapsed the

L2 and L4 LES recordings as they did not significantly

differ. Fig. 4 shows the results of the PCA on LES activity of

subsequent stride cycles for the first three PCs. The

contribution of the LBP group to PC1 was significantly

smaller than that of the control group for LES left

(F (1,46) = 5.09, P = 0.03) and right (F (1,46) = 11.95,

P < 0.001). In addition, there was a significant HS � ve-

velocity interaction for left (F (5,230) = 5.8, P < 0.01) and

right (F (5,230) = 2.3, P = 0.04) LES since the velocity-

induced changes were smaller in the LBP than in the control

Fig. 4. Projections j(1,. . .,3) of the first three principal components (left panel) and

corresponding pattern. PC1 represented peak in LES activity around foot contact, P

PC3 captured modifications due to velocity changes. Dotted vertical lines indica

averaged across subjects.

group. Also, the fine-tuning of LES activity to velocity

changes represented by PC2 was significantly smaller in the

LBP group for left (F (1,46) = 4.9, P = 0.03) and right

(F (1,46) = 4.7, P = 0.04) LES. We found a significant

HS � velocity effect on right and left LES activity (both

F (5,230) = 3.4, P < 0.01) for the PC3 coefficients. In the

control group, peak activity occurred later at low velocities

than at higher ones. Conversely, in the LBP group

coefficients were larger at higher than at lower velocities

because peak activity occurred earlier at higher speed. The

remaining PC4 to PC6 primarily represented modifications

of the pattern due to the rather erratic LES activity at 1.4 and

2.2 km/h without significant differences between groups.

The variability of global and residual patterns of left and

right LES activity differed significantly between LBP and

control group (see Table 3). Due to the smaller contribution

of the LBP group to the global pattern, the variability was

lower at all walking velocities in the LBP group, with ratios

smaller than 1 (except for the 1.4 km/h velocity). In addition,

variability of the residual pattern of LES activity was higher

at all velocities and decreased less with increasing velocities

in the LBP than in the control group (Fig. 5).

the contribution of left and right LES of the control and LBP group to the

C2 reflected modifications of the step frequency around double support, and

te moments of right foot contact. Eigenvector coefficients v(1,. . .,3) are here

C.J.C. Lamoth et al. / Gait & Posture 23 (2006) 230–239 237

Table 3

Significant effects of walking velocity (Vel), and HS (control vs. LBP group) on the variability of the global and residual pattern of left and right lumbar erector

spinae (LES) activity, and significant effects of post-hoc analyses

Variability LES ANOVA Post-hoc test HS

Vel Vel � HS HS 6.2 km/h 5.4 km/h 2.2 km/h 4.6 km/h

F(5,230) P-value F(5,230) P-value F(1,22) P-value t(46) P-value t(46) P-value t(46) P-value t(46) P-value

Global pattern

Left LES 41.0 <0.01 2.5 0.03 4.8 0.03 2.4 0.02 3.2 <0.01 2.8 0.01

Right LES 28.2 <0.01 3.6 <0.01 8.8 <0.01 2.8 0.01 3.9 <0.01 2.1 0.04 4.3 <0.01

Residual pattern

Left LES 41.4 <0.01 3.4 0.01 6.2 0.02 �2.5 0.02 �3.0 <0.01 �2.1 0.04 �2.7 0.01

Right LES 29.5 <0.01 5.7 <0.01 10.1 <0.01 �3.3 <0.01 �4.3 <0.01 �2.2 0.03 �4.3 <0.01

4. Discussion

The present experiment examined the capacity of

persons with LBP to adapt their gait patterns to sudden

changes in walking velocity. Invariant and variant proper-

ties of trunk–pelvis coordination and accompanying LES

activity revealed that, unlike healthy persons, they had

difficulties to adapt trunk–pelvis coordination flexibly and

LES activity to perturbations in velocity. In line with

previous studies [17,19], a velocity-induced change of

thorax–pelvis coordination toward antiphase was found to

be more pronounced in the controls than in the LBP

subjects. The latter also tended to move their lumbar and

pelvic segments as a rigid unit. The change of variability in

coordination patterns caused by velocity perturbations

differed markedly between LBP and control subjects.

Following large velocity perturbations the variability of

Fig. 5. Variability of global (left panel) and residual patterns (right panel) of left an

cycles for LBP subjects over that of the control group (ratio). Error bars indicat

both thoracic–pelvic and lumbar–pelvic coordination in

the transverse plane was strongly reduced in the LBP

subjects, whereas in the frontal plane intersegmental

coordination was more variable and less tightly coupled.

PCA revealed that variability of residual patterns of

thoracic and lumbar rotations was reduced in the transverse

and enhanced in the frontal plane, particularly at higher

velocities. Irrespective of velocity, the variability of pelvic

rotations was also enhanced in LBP subjects. Most likely,

LBP subjects were unable to accustom themselves to a new

velocity level within the available time. The increased

trunk variability in the frontal plane hinted at an effort of

LBP subjects to actively control overall gait stability in the

absence of antiphase thorax–pelvis coordination at higher

velocities. Note that the variability of the residual patterns

was not only random but also included deviations from the

global pattern, e.g. higher harmonics.

d right LES activity quantified by calculating the average variance over stride

e S.E.

C.J.C. Lamoth et al. / Gait & Posture 23 (2006) 230–239238

Significant differences between the contribution of

healthy and LBP subjects to the global pattern of

amplitude-normalised or non-normalised kinematic data

were not present. The intersegmental coordination was

mainly disturbed in the LBP subjects in the absence of

significant differences in the amplitudes of the rotations. In

line we observed considerable differences between groups in

LES activity. PCA indicated that muscular control in LBP

was affected by alterations of the global pattern, a

diminished capacity to adapt LES activity to changing

velocities, and a marked increase in variability of the

residual pattern, particularly at higher velocities. In the LBP

subjects, the residual pattern consisted of rather irregular

deviations of the global pattern in terms of phase shifts,

amplitude modifications, and additional bursts of activity.

These changes in LES activity showed poor control of LES

muscle activity due to LBP.

Changes in LES activity in LBP may reflect an attempt to

stabilise the spine (by increasing the stiffness) in the face of

potentially disruptive perturbations, although they may also

reflect a reduced proprioception due to persisting complaints

[23,24]—both possibilities require further consideration.

Indeed, spinal instability from dysfunction has been

emphasized as an important aspect of LBP [25]. Spinal

stability can be affected through changes in spinal structures,

reduced control over trunk muscles, and changes in

proprioception. All these changes may be both a cause

and an effect of injury [26] and may contribute to LBP

becoming chronic but, to date, the mechanisms are

unknown. The smaller relative phases and the reduced

variability in the transverse plane, particularly after large

velocity perturbations, might be seen as evidence for

increased stiffness of the spine. This could also explain the

limited ability of persons with LBP to adjust trunk–pelvis

coordination to velocity changes. Alternatively the observed

changes in LES activity may reflect altered proprioception in

LBP and may play a role besides spinal instability. Persons

with chronic LBP may have reduced lumbar position sense

[27] and poor balance while standing [28]. The reception of

less, or possibly inappropriate, spatial, temporal, or kinetic

information might be important for the precise timing of

trunk and pelvis rotations during walking.

5. Conclusion

The present study supports the hypothesis that persons

with chronic LBP have limitations in motor control.

Locomotor problems in persons with LBP were primarily

coordinative in nature. An inability to adapt invariant and

variant properties of gait patterns to changes in velocity can

reduce overall gait stability. In this sense, the slower natural

walking velocity of persons with LBP becomes a functional

adaptation since it may enable persons with LBP to cope

with internal and external perturbations. Such an adaptation

has also considerable disadvantages because it restricts

possible velocity alterations required to achieve behavioural

goals (catching the bus, avoid a cyclist, etc.). This implies

that conservative management of chronic LBP patients is

probably most effective when it includes techniques that are

aimed at promoting trunk, pelvis, muscle coordination and

postural control, with the ultimate goal to improve

functional capacity and flexibility.

Acknowledgements

This study was supported by grants from the Dutch

organization for Scientific Research #904-65-090 (MW-

NWO), the Dutch Association for Exercise Therapy

Mensendieck (NVOM) and the Mensendieck Development

Foundation (SOM).

References

[1] Khodadadeh S, Eisenstein SM, Summers B, Patrick J. Gait asymmetry

in patients with chronic low back pain. Neuro-Orthop 1988;6:24–7.

[2] Keefe FJ, Hill RW. An objective approach to quantifying pain behavior

and gait patterns in low back pain patients. Pain 1985;12:153–61.

[3] Spenkelink CD, Hutten MM, Hermens HJ, Greitemann BO. Assess-

ment of activities of daily living with an ambulatory monitoring

system: a comparative study in patients with chronic low back pain

and nonsymptomatic controls. Clin Rehabil 2002;16:16–26.

[4] Buzzi UH, Stergiou N, Kurz MJ, Hageman PA, Heidel J. Nonlinear

dynamics indicates aging affects variability during gait. Clin Biomech

2003;18:435–43.

[5] Dingwell JB, Cavanagh PR. Increased variability of continuous over-

ground walking in neuropathic patients is only indirectly related to

sensory loss. Gait Posture 2001;14:1–10.

[6] Kwakkel G, Wagenaar RC. Effect of duration of upper- and lower-

extremity rehabilitation sessions and walking speed on recovery of

interlimb coordination in hemiplegic gait. Phys Ther 2002;82:433–48.

[7] Wu W, Meijer OG, Jutte PC, Uegaki K, Lamoth CJC, de Wolf GS, van

Dieen JH. Gait in patients with pregnancy-related pain in the pelvis: an

emphasis on coordination of transverse pelvic and thoracic rotations.

Clin Biomech 2002;17:678–86.

[8] Beek PJ, Peper CE, Stegeman D. Dynamical models of movement

coordination. Hum Mov Sci 1995;14:573–608.

[9] Daffertshofer A, Lamoth CJC, Meijer OG, Beek PJ. PCA in studying

coordination and variability: a tutorial. Clin Biomech 2004;19:415–

28.

[10] Donker SF, Beek PJ, Wagenaar RC, Mulder T. Coordination between

arm and leg movements during locomotion. J Motor Behav

2001;33:86–102.

[11] Lamoth CJC, Beek PJ, Meijer OG. Pelvis–thorax coordination in the

transverse plane during gait. Gait Posture 2002;16:101–14.

[12] Wannier T, Bastiaanse C, Colombo G, Dietz V. Arm to leg coordina-

tion in humans during walking, creeping and swimming activities. Exp

Brain Res 2001;141:375–9.

[13] Yamasaki T, Nomura T, Sato S. Phase reset and dynamic stability

during human gait. Biosystems 2003;71:221–32.

[14] Crosbie J, Vachalathiti R, Smith R. Patterns of spinal motion during

walking. Gait Posture 1997;5:6–12.

[15] Lamoth CJC, Daffertshofer A, Meijer OG, Moseley G, Wuisman

PIJM, Beek PJ. Effects of experimentally induced pain and fear of

pain on trunk coordination and back muscle activity during walking.

Clin Biomech 2004;19:551–63.

C.J.C. Lamoth et al. / Gait & Posture 23 (2006) 230–239 239

[16] Stokes VP, Andersson CA, Forssberg H. Rotational and translational

movement features of the pelvis and thorax during adult human

locomotion. J Biomech 1989;22:43–50.

[17] Lamoth CJC, Meijer OG, Daffertshofer A, Wuisman PIJM, Beek PJ.

Effects of chronic low back pain on trunk coordination and back

muscle activity during walking: changes in motor control. Eur Spine J,

in press.

[18] Thorstensson A, Carlson H, Zomlefer MR, Nilsson J. Lumbar back

muscle activity in relation to trunk movements during locomotion in

man. Acta Physiol Scand 1982;116:13–20.

[19] Lamoth CJC, Meijer OG, Wuisman PIJM, van Dieen J, Levin MF,

Beek PJ. Pelvis thorax coordination in the transverse plane during

walking in persons with nonspecific low back pain. Spine 2002;27:

E92–9.

[20] Vink P, Daanen HAM. Verbout AJ. Specificity of surface-EMG on the

intrinsic lumbar back muscles. Hum Mov Sci 1989;8:67–78.

[21] Pijnappels M, Bobbert MF, van Dieen JH. Changes in walking pattern

caused by the possibility of a tripping reaction. Gait Posture

2001;14:11–8.

[22] Fisher NI. Statistical analysis of circular data. Cambridge: Cambridge

University Press, 1993.

[23] Van Dieen JH, Cholewicki J, Radebold A. Trunk muscle recruitment

patterns in low back pain patients enhance the stability of the lumbar

spine. Spine 2003;28:834–41.

[24] Van Dieen JH, Selen LP, Cholewicki J. Trunk muscle activation in low-

back pain patients, an analysis of the literature. J Electromyogr

Kinesiol 2003;13:333–51.

[25] Panjabi MM. The stabilizing system of the spine. Part I. Function,

dysfunction and enhancement. J Spinal Disord 1992;5:383–9.

[26] Parkhurst TM, Burnett CN. Injury and proprioception in the lower

back. J Orthop Sports Phys Ther 1994;19:282–95.

[27] Gill KP, Callaghan MJ. The measurement of lumbar proprioception

in individuals with and without low back pain. Spine 1998;23:

371–7.

[28] Luoto S, Aalto H, Taimela S, Hurri H, Pyykko I, Alaranta H. One-

footed and externally disturbed two-footed postural control in patients

with chronic low back pain and healthy control subjects. A controlled

study with follow-up. Spine 1998;23:2081–90.