Anda di halaman 1dari 12

SPINE Volume 29, Number 11, pp 12541265

2004, Lippincott Williams & Wilkins, Inc.


Determining the Stabilizing Role of Individual Torso
Muscles During Rehabilitation Exercises
Natasa Kavcic, MSc, Sylvain Grenier, PhD, and Stuart M. McGill, PhD
Study Design. A systematic biomechanical analysis
involving an articial perturbation applied to individual
lumbar muscles in order to assess their potential stabiliz-
ing role.
Objectives. To identify which torso muscles stabilize
the spine during different loading conditions and to iden-
tify possible mechanisms of function.
Summary of Background Data. Stabilization exercises
are thought to train muscle patterns that ensure spine
stability; however, little quantication and no consensus
exists as to which muscles contribute to stability.
Methods. Spine kinematics, external forces, and 14
channels of torso electromyography were recorded for
seven stabilization exercises in order to capture the indi-
vidual motor control strategies adopted by different peo-
ple. Data were input into a detailed model of the lumbar
spine to quantify spine joint forces and stability. The EMG
signal for a particular muscle was replaced either unilat-
erally or bilaterally by a sinusoid, and the resultant
change in the stability index was quantied.
Results. A direction-dependent-stabilizing role was
noticed in the larger, multisegmental muscles, whereas a
specic subtle efciency to generate stability was ob-
served for the smaller, intersegmental spinal muscles.
Conclusions. No single muscle dominated in the en-
hancement of spine stability, and their individual roles
were continuously changing across tasks. Clinically, if the
goal is to train for stability, enhancing motor patterns that
incorporate many muscles rather than targeting just a few
is justiable. [Key words: lumbar spine, spine stability,
modeling, muscles] Spine 2004;29:12541265
While muscles function to create torques, which support
postures and facilitate movement, they are also critical
for ensuring spine stability.
1
Clinically, the question of
how to train lumbar spine stability requires knowledge
of howthe various muscles contribute to ensuring stabil-
ity. A common functional distinction used to classify the
role of the different muscles is that intersegmental or
local muscles are hypothesized to function primarily
as stabilizers and multisegmental or global muscles are
hypothesized to function primarily as moment produc-
ers.
24
This distinction, formalized by Professor Anders
Bergmark,
2
focused the early discussions on stability;
however, debate continues over which muscles are im-
portant stabilizers and how to best train the neuromus-
cular control system to ensure sufcient stability.
Researchers have used various techniques to investi-
gate the question of which muscles stabilize the lumbar
spine. Electromyographic analysis of torso muscle onset
times to various perturbations has suggested that the
more internal muscles, particularly the transverse abdo-
minis and internal obliques, behave in an anticipatory
manner, irrespective of loading condition, suggesting a
proactive control of spine stability.
5,6
Others have ob-
served a wasting of the multidus muscle on the side of
the reported low back pain with MRI techniques,
7
sug-
gesting that in order to ensure a stable spine this muscle
requires specic training to return the cross-sectional
area of multidus to normal levels.
4
While these studies
did not quantify stability but rather relied on qualitative
intuition, other approaches have attempted to quantify
stability. For example, in vitro approaches have repre-
sented muscle forces with wire cables acting on cadaveric
lumbar spines. Through investigation of predominantly
small, local muscles, several researchers have found
that these muscles successfully increase the stiffness
within the spinal structure, critical for stability.
810
While many believe that the local muscles are cru-
cial for spine stability, others hypothesize that the
global, larger muscles play a role. Panjabi et al
8
sug-
gested that the role global muscles have in stabilizing the
lumbar spine comes from their efcient ability to impact
the stiffness of the entire spinal column, opposed to local
muscles that can only act on a fewjoints. Cholewicki and
McGill
11
and Cholewicki and Van Vliet
12
suggest from
the results of their biomechanical analyses, that no single
muscle, local or global, possesses a dominant responsi-
bility for lumbar spine stability and therefore concluded
that training efforts should not focus on any single
muscle.
Contrasting results and descriptions for the neuro-
muscular control of spine stability have led to the devel-
opment of various training theories. Patterns such as an-
tagonist cocontraction as a method of increasing spine
stiffness has been conrmed through numerous stud-
ies
1316
; however, some argue that enhancing this re-
sponse for therapeutic purposes to train spine stability
can lead to very high compressive load penalties.
4
Some
advocate training isolated groups of muscles, primarily
local, with the goal to minimize global muscle activa-
tion and compressive loads. Identifying muscle impor-
From the University of Waterloo, Faculty of Applied Health Sciences,
Waterloo, Ontario, Canada.
Acknowledgment date: May 21, 2003. First revision date: June 26,
2003. Acceptance date: August 6, 2003.
Supported by the National Science and Engineering Research Council
of Canada.
The manuscript submitted does not contain information about medical
device(s)/drug(s).
Federal funds were received in support of this work. No benets in any
form have been or will be received from a commercial party related
directly or indirectly to the subject of this manuscript.
Address correspondence to Stuart M. McGill, PhD, Faculty of Applied
Health Science, University of Waterloo, Waterloo, Ontario, Canada
N2L 3G1; E-mail: mcgill@healthy.uwaterloo.ca
1254
tance during different loading conditions is necessary to
properly critique existing clinical practices for training
and restoring a healthy lumbar spine.
The purpose of this study is to use a highly sophisti-
cated and detailed torso model, driven with biologic sig-
nals measured directly from each study participant, to
compute spine loads and stability. Each muscle was sys-
tematically adjusted to assess the impact on the stability
of the spine, thereby quantifying their contribution at a
specic instant in time. Furthermore, an attempt to iden-
tify the different mechanisms as to how the various mus-
cles contribute to lumbar spine stability was performed.
Materials and Methods
Ten male study participants performed a series of eight differ-
ent exercises (Figure 1) while electromyography, three-
dimensional lumbar motion, and external forces were mea-
sured. These data were input into a series of biomechanical
models in order to calculate a measure of lumbar joint forces
and spine stability. These methods are extremely detailed and
have already been published. While the interested reader can
refer to the manuscripts for details,
11,17,18
the essential details
are documented here. A schematic of the protocol is shown in
Figure 2. All procedures were approved by the University Ofce
for Research Ethics.
Subjects. Ten male university students with an average age of
21 years (SD 3 years), height of 177.8 cm (SD 6.2 cm), and
weight of 80.2 kg (SD 12.1 kg) volunteered to participate in
this study. Subjects had no history of low back pain. Before
testing, study participants height, weight, and breadth dimen-
sions at the feet, ankles, knees, hips, hands, wrists, elbows, and
shoulders were obtained while standing in anatomic position.
Figure 1. Pictures of different stabilization exercises. A, Abdominal curl. B, Right side bridge. C, Sitting on a stool. D, Sitting on a gym ball.
E, Four-point kneeling with contralateral arm and leg extension. F, Four-point kneeling with single leg extension. G, Back bridge with single
leg extension. H, Back bridge.
1255 Torso Muscles and Rehabilitation Exercises

Kavcic et al
Data Collection
Exercises. Each study participant performed a series of eight
exercises presented in random order. The exercises (shown in
Figure 1) include the abdominal curl (A), right side bridge (B),
sitting on a gym ball (D), four-point kneeling with a left arm
and right leg extension (E), four-point kneeling with right leg
extension (F), back bridging with right leg extension (G), and
back bridging (H). To act as a control trial for the gym ball
condition and allow for assessment of unstable support sur-
faces, study participants performed trials sitting on a stool (C).
Figure 2. Flow chart of the various models used in the stability analysis.
1256 Spine

Volume 29

Number 11

2004
The stabilization exercises were chosen for ease of analysis as
well as to ensure moments to the spine in all three axes of
rotation (Table 1). Each exercise was performed with a neutral
lumbar spine position and controlled limb positioning. Limb
and/or pelvis position was controlled through the use of an
external frame with metal bars that was placed alongside body
segments to act as targets.
Each exercise was held isometrically for 2 seconds with an
isometric contraction of the abdominal muscles (termed ab-
dominal brace).
1
A brace is an isometric contraction of all the
muscles of the abdominal wall without any change in the po-
sition of the muscles. This is in contrast to the abdominal
hollow, described by Richardson et al,
4
which is intended to
focus on the recruitment of the transverse abdominis while
minimizing activation of the rectus abdominis and the ob-
liques. Bracing has been shown to be superior to hollowing for
enhancing lumbar stability.
23
Subjects were shown the tech-
nique for performing both the hollow and brace abdom-
inal contraction, and the instruction was to perform the
brace to the same perceived intensity as a hollow. Conse-
quently, the intensity of the contraction was fairly low, however,
since the intensity was chosen subjectively, the stability demand of
some postures may have required that study participants brace
more intensively than originally instructed. Subjects were given an
unlimited number of practice trials and once comfortable with the
technique of performing each exercise with an abdominal brace,
three successive trials were measured.
Instrumentation
Electromyography. Fourteen channels of EMG were col-
lected fromthe following muscles bilaterally: rectus abdominis,
internal oblique, external oblique, latissimus dorsi, thoracic
erector spinae (longissimus thoracis and iliocostalis at T9),
lumbar erector spinae (longissimus and iliocostalis at L3), and
multidus (1cmlateral to L5). We acknowledge the difculty
in capturing multidus with surface electrodes
19
and therefore
assign validity of the EMG signal to the landmarked location
rather than to the multidus muscle itself. Ag-AgCl surface
electrodes were positioned with an interelectrode distance of
about 3 cm. The EMG signals were amplied and then A/D
converted with a 12-bit, 16-channel A/D converter at 1,024
Hz. Each study participant was required to perform a maximal
contraction of each measured muscle for normalization of each
channel. For the abdominal muscles each study participant,
while in a sit-up position and manually braced by a research
assistant, produced a maximal isometric exor moment fol-
lowed sequentially by a right and left lateral bend moment and
then a right and left twist moment; little motion took place. For
the extensor muscles, a resisted maximum extension in the
Biering-Srensen position was performed with focus on quasi-
static motion throughout neutral lordosis, which was found to
create larger neural drive. The EMG signal was normalized to
these maximal contractions, full wave rectied and low-pass
ltered with a second-order Butterworth lter. A cutoff fre-
quency of 2.5 Hz was used to mimic the frequency response of
the torso muscles.
11
Three-Dimensional Kinematic Positioning of the Lumbar
Spine. Lumbar spine kinematics was measured about three or-
thogonal axes using a 3 Space IsoTRAK, electromagnetic
tracking instrument (Polhemus Inc., Colchester, VT). This in-
strument consists of a single transmitter that was strapped to
the pelvis over the sacrum and a receiver strapped across the
ribcage, over the T12 spinous process. Thus, the position of the
ribcage relative to the sacrum was measured, isolating lumbar
motion. Overall rotation of the lumbar spine was normalized
relative to each study participants standing neutral spine pos-
ture. In this way, individual variance in the passive tissue con-
tributions as a function of maximum range of motion was
represented. However, in this experiment, there was minimal
contribution of the passive tissue restorative moment because
of the neutral spine posture characteristic of the stabilization
exercises chosen.
External Force Measures. For exercises requiring an inverse
dynamic load application, namely, the four-point kneeling ex-
ercises, back bridging exercises, and the side bridge, external
force measures were recorded using an AMTI force plate. The
signals were amplied to produce a peak to peak range of 20 V
(10V) and then A/Dconverted with a 12-bit A/Dconverter at
1,024 Hz. Forces and moments were measured about three
axes and were used to calculate the external force center of
pressure values in the x, y, and z direction. For each exercise,
the study participant was instructed to position the contacting
segment on the force plate around the 0, 0, 0-reference point
located at the center of the force plate. Reaction forces were
measured at different parts of the upper body depending on the
exercise being performed (Figure 1). Force plate measures were
not recorded for the abdominal curl or ball sitting and chair
sitting exercises. The process of using whole body linked seg-
ment dynamics and measured external forces has been ex-
plained previously.
11
Kinematic Limb Positions. Kinematic marker data for each
exercise were measured from a single study participant, not
part of the group of 10 mentioned above. This study partici-
pant had a height of 178 cmand a weight of 79 kg. The external
segment kinematics were recorded for each exercise posture
with a single digital video image and guided by a space frame
jig. The isometric position of each exercise was used to analyze
the segment kinematics in the sagittal plane. The joints digi-
tized for the kinematic analysis were the metatarsal, ankle, hip,
shoulder, elbow, wrist and hand bilaterally, as well as L4L5
and C7T1. The kinematic posture obtained for each exercise
was controlled in the other 10 study participants with the ex-
ternal jig, and the marker data were scaled to the height of each
individual study participant. The joint locations about the z-
axis, or in the frontal plane, were scaled to the breadth mea-
Table 1. Summary of the support moments created at the
L4L5 joint in order to perform the different exercises
Average L4L5 moment (Nm)
Bend ( 1 SD) Twist ( 1 SD) Flex ( 1 SD)
Abdcurl 1.30 (1.9) 0.72 (0.99) 56.71 (7.0)
Chair 0.54 (0.5) 0.10 (0.3) 1.47 (0.5)
Ball 0.72 (1.0) 0.18 (0.5) 1.28 (0.5)
Bridge 0.15 (3.9) 2.64 (7.6) 73.81 (32.7)
Bridge leg 8.42 (5.0) 15.74 (7.6) 65.94 (33.3)
Fpn_leg 4.84 (2.9) 15.62 (8.1) 6.14 (25.3)
Fpn_arm/leg 0.05 (5.1) 57.05 (14.6) 32.84 (23.2)
Side bridge 69.18 (21.9) 12.80 (3.9) 2.87 (3.4)
Average and standard deviations are listed. In the sagittal plane, exion is
negative and extension is positive. In the frontal plane, right lateral bend is
positive and left lateral bend is negative. In the transverse plane, right axial
twist is negative and left axial twist is positive.
1257 Torso Muscles and Rehabilitation Exercises

Kavcic et al
sures taken from each study participant. Since no exercise re-
quired deviations of the limbs from anatomic position in the
frontal plane, breadth measures were assumed to be constant
across exercises.
Data Analysis
Calculating a Stability Index. The analysis of stability was
performed using a method documented by Cholewicki and
McGill
11
and involved three cascading and interdependent
models. For the interested reader, these models are described in
detail by Cholewicki and McGill,
11
McGill and Norman,
17
and McGill
18
; however, a brief description is provided here
(Figure 2, ow chart showing the modeling process for the
stability analysis).
The rst model is an 8-segment link segment model that uses
external force measures recorded from the forceplate, study
participant kinematics, and anthropometrics of height and
weight to calculate reaction forces and moments acting at each
of the 6 lumbar intervertebral joints through a top-down, in-
verse dynamics approach. The L4L5 moments calculated
fromthis linked-segment model are used to ultimately drive the
EMG-assisted optimization routine that determines the muscle
force proles; however, this will be described in more detail
later in this section.
20
The reaction forces fromthe link segment
model calculations are used to determine the shear and com-
pression forces at the L4L5 joint.
The second model is the lumbar spine model, which con-
sists of an anatomically detailed, three-dimensional ribcage,
pelvis/sacrum, and ve intervening vertebrae. More than 100
laminae of muscle and the passive tissues, which are repre-
sented as a lumped parameter of torsional stiffness, are mod-
eled about each axis. This model uses the measured three-
dimensional relative spine motion data from the 3-space
IsoTRAK system and assigns the appropriate rotation to each
of the lumbar vertebral segments based on ndings from White
and Panjabi.
21
Muscle lengths and velocities are determined from
their motions and attachment points on the dynamic skeleton of
which the motion is driven from the directly measured lumbar
kinematics obtained from the study participant. As well, the ori-
entation of the vertebral segments along with stress/strain rela-
tionships of the passive tissues was usedtocalculate the restorative
moment created by the spinal ligaments and discs.
The third model, termed the distribution-moment mod-
el,
22
is used to calculate the muscle force and stiffness proles
for each of the muscles. The model uses the normalized EMG
prole of each muscle along with the calculated values of mus-
cle length and velocity of contraction to calculate the active
muscle force and any passive contribution from the parallel
elastic components. When input to the spine model, these mus-
cle forces are used to calculate a moment for each of the 18 df
of the six intervertebral joints. The objective function for the
EMG-assisted optimization routine is to match the moments
with a minimal amount of change to the EMG driven force
proles. In this way, biologic validity of using EMG is pre-
served while mathematical validity is addressed with achieving
balanced moments. The adjusted muscle force and stiffness
proles are then used in the calculations of L4L5 compression
and shear, as well as in calculating spine stability. The most
recent updates to the model, specically regarding the much
improved representation of the transverse abdominis, are doc-
umented by Grenier and McGill.
23
The value for stability, or stability index, was obtained by
calculating a level of potential energy in the spinal structure for
each of the 18 df (three rotational axes at six lumbar joints)
resulting from the combined potential energy existing in both
the active and passive spinal structures, minus any work done
from external loads. The 18 values of potential energy were
formed into an 18 18 Hessian matrix and diagonalized. The
determinant of this matrix represented an index of spine stability.
For a more detailed description of the mathematical procedures,
refer to Cholewicki and McGill
11
and for sensitivity testing and
mathematical validity of the approach see Howarth et al.
24
Before inputting data into the link-segment model, certain
modications were made to both the data and the model so to
enable accurate calculations of spine load and stability for cer-
tain exercise postures. They are noted as follows:
Abdominal curl. When performing this exercise, study par-
ticipants were directed to perform a curl-up such that rotation
of the upper body occurred about the base of the rib cage.
Consequently, the weight supported consisted of the head and
neck, thorax, and arms. Calculating moments about the L4L5
joint would consider the entire torso mass and result in an
overestimation of the exor moment required by the muscles.
To consider the true axis of rotation, the L4L5 marker was
shifted up along the long axis of the spine to accurately repre-
sent a rotation of the thorax opposed to the trunk. A thorax
distance of 0.4 m, which is characteristic of a 75 percentile male
was used. The mass proportion assigned to the thorax was
0.216 of body mass.
25
For this exercise only, the abdomen
segment was considered a rigid segment and the thorax mo-
ment was then translated to the L4L5 joint, recognizing that
the rectus abdominis carries equal loading along its length.
Bridging with single leg extension. In this exercise, the inter-
nal oblique activation prole did not accurately represent that
of the psoas muscle because of the extended leg. To account for
the extra force necessary to support the extended leg, the psoas
force in the lifted leg was calculated as a proportion of the
moment supporting the leg, which was assumed to be primarily
generated from combined action of the rectus femoris, iliacus,
and psoas. The moment arms and peak isometric muscle forces
used to calculate the proportions for the three listed muscles
were obtained fromthe literature (Table 2).
26,27
Then, for each
study participant, the support moment required to maintain
the posture of the lifted leg was calculated. This moment was
then multiplied by a proportionality constant for psoas and
divided by its moment arm. The resulting force value was input
into the 18 df lumbar spine model (Figure 2) by adding it
directly to the compressive force acting on the spine, consistent
with the psoas line of action.
28
Table 2. Parameters used to calculate the contribution
of the psoas muscle to the support moment of the
extended leg for the back bridge with single leg
extension
Muscle
Peak isometric
muscle force (N)*
Moment
arm (cm)
Relative proportion of
total hip-exion
moment
Psoas 370 2.9 0.19
Iliacus 430 3.0 0.23
Rectus Femoris 780 4.2 0.58
* From Delp et al.
26
Moment arms are measured at the hip during the mid stance phase of gait.
Arnold et al.
27
1258 Spine

Volume 29

Number 11

2004
Determining a Muscles Impact on Spine Stability. The
contribution of each individual muscle to spine stability was
evaluated in the following way. A value of external force and
muscle activation for each muscle was taken at a point in time
corresponding to the 1-second point of each 2-second trial.
This prole was then frozen and extended for the entire
duration of the 2-second trial. The activation prole for all
fascicles of a single muscle of interest, or target muscle, was
then replaced by a sinusoid wave that varied from0%to 100%
MVC (Figure 3A), and the analysis was run with the new mus-
cle prole. A sinusoid wave was originally chosen as an input
because it was thought that some muscles would demonstrate
more control over stability than others. Control of spine sta-
bility would be reected by a strong correlation between
changes in stability with changes in muscle activation. In those
muscles that had little control over spine stability, the sinusoi-
dal pattern would be less evident in the stability output. A
sinusoid was chosen as a very specic input that could be iden-
tied in the output. Through a pilot analysis, however, there
appeared to be no signicant difference in how closely stability
followed muscle activation across the various muscles tested;
therefore, this analysis was not performed.
The specic target muscles assessed were the rectus abdomi-
nis, external oblique, internal oblique, pars lumborumbers of
longissimus thoracis and iliocostalis lumborum, thoracic bers
of iliocostalis lumborum, longissimus thoracis, quadratus lum-
borum, latissimus dorsi, multidus, and transverse abdominis.
This analysis was systematically repeated for each muscle, one
at a time, both unilaterally and bilaterally. To isolate the effect
of each target single muscle at this level of analysis, the EMG-
assisted optimization routine (Figure 2), used to balance the
moments, was not used. This prevented the force and stiffness
proles of the other muscles from changing. In effect, this pro-
Figure 3. Sinusoidal muscle acti-
vation prole from 0 to 100%
MVC (A) and the associated
change in the stability index
when manipulating each muscle
EMG prole (B). RL corresponds
to the muscle on both the right
and left side. Rect rectus ab-
dominis; Ext external oblique;
Int internal oblique; Pars
pars lumborum bers of longissi-
mus thoracis and iliocostalis
lumborum; Ilio thoracic bers
of iliocostalis lumborum; Long
thoracic bers of longissimus
thoracis; Quad quadratus lum-
borum; Lat latissimus dorsi;
Mult multidus; Trans
transversus abdominis.
1259 Torso Muscles and Rehabilitation Exercises

Kavcic et al
cedure allowed each muscle to be rattled and the subsequent
effect of this perturbation on the spine assessed.
The articial sinusoidal activation prole impacted many
variables within the analysis; however, the effect was only
quantied in certain variables of interest: namely, muscle force,
muscle stiffness, spine loads, and the stability index. The max-
imum increase and decrease in these variables, resulting from
the sinusoidal manipulation, were computed and compared to
a nonmanipulated control trial.
In an attempt to better understand the different mechanical
advantages for each of the muscles to stabilize, the RMS differ-
ence was calculated across each stability index curve resulting
from the manipulated muscle activation prole, as well as
across the corresponding muscle force curve. The RMS differ-
ence was used to quantify of the magnitude of uctuation
within the particular curve. The force RMS difference was then
divided into the stability index RMS difference. In this sense, an
efciency ratio was created to describe the coupling between
the uctuations in the force of a particular muscle and the
corresponding uctuations in spine stability.
Results
The effect of the sinusoidal EMGactivation prole on the
calculated stability index is shown for each muscle in
Figure 3B The stability index for each manipulated mus-
cle is superimposed on the same graph.
Assessing the Absolute Impact of a Single Muscle on
Lumbar Spine Stability
The effect of increasing each muscle activation prole to
100%MVCon increasing the stability index is shown in
Figure 4A, whereas the effect from decreasing muscle
activation to 0% MVC is shown in Figure 4B. A major
Figure 4. A, Increase in stability
index resulting from activating a
muscle bilaterally to 100% MVC.
B, Decrease in stability index re-
sulting from turning a muscle off
bilaterally to 0% MVC. Across
tasks there is no consistent pat-
tern in the ability of the different
muscles to affect stability. How-
ever, it appears as though,
across the larger muscles, in-
creased activation of the mo-
ment antagonist enhances stabil-
ity and decreased activation of
the moment agonist reduces
stability.
1260 Spine

Volume 29

Number 11

2004
nding in the analysis is that, between the different tasks,
there is no consistent pattern across muscles in their abil-
ity to affect stability. This is particularly evident with
some of the larger muscles, such as the rectus abdominis
and the lumbar and thoracic extensors. In contrast, the
quadratus lumborum, latissimus dorsi, multidus, and
transverse abdominis demonstrated only small changes
in their relative patterns in terms of both increasing and
decreasing stability.
Quantication of each muscles absolute impact on
the stability index shows that, compared with the rectus
abdominis, obliques, and lumbar and thoracic extensors,
the quadratus lumborum, latissimus dorsi, multidus,
and transverse abdominis each created minimal changes.
In contrast, both the internal and external obliques con-
sistently demonstrated a large impact on both increasing
and decreasing stability irrespective of the task condi-
tion. Between the two muscles, a more dramatic effect
was produced fromthe internal obliques. One important
note is that in the stabilization exercises assessed, no
individual muscle, either unilaterally or bilaterally, when
articially reduced in activation, created an unstable
situation.
An interesting result is that certain muscles demon-
strated a direction-dependent effect on lumbar spine sta-
bility. Specically, coactivation of what would be con-
sidered an antagonist, in a torque context, enhances
stability. Compare the rank order of the predominant
exor: rectus abdominis versus the major lumbar exten-
sors, pars lumborum, iliocostalis lumborum, and longis-
simus thoracis. In the abdominal curl, which is a exion
dominant task (Table 1), the three extensor muscles
demonstrate a greater effect on increasing the stability
index compared with the rectus abdominis. However,
this pattern is reversed when quantifying each muscles
ability to reduce spine stability. In contrast, during the
extension dominant tasks (Table 1), the rectus abdomi-
nis creates a greater increase in stability over the pars
lumborum and longissimus thoracis. For the iliocostalis,
lumborum, this pattern is not so evident; however, care-
ful examination shows that across the extension domi-
nant tasks, as the required support moment increases,
the relative difference between the effects of the rectus
abdominis and the iliocostalis lumborum decreases. As
with the abdominal curl, when the activation levels are
reduced to 0% MVC, the pattern between the exors
and extensors reverses.
The same association observed between the exor and
extensor muscle groups is observed between certain right
versus left muscle groups during asymmetric tasks such as
the four-point kneeling tasks and side bridge (Figure 5).
It should be noted that the above results refer to group
means calculated from the 10 study participants. Across
the individual study participants, the pattern of muscle
impact on stability was not consistent for any given task;
however, the direction-dependent effect observed among
the group means exists at an individual level as well.
Assessment of a Potential Mechanical Stabilizing
Mechanism for Each of the Different Muscles
The cost of certain muscles to stabilize is demonstrated in
Figure 6. These gures showthat the larger muscles, such
as the rectus abdominis, obliques, and upper and lower
erectors, impose larger changes in L4L5 load compared
with the other muscles tested. Given this, the nal anal-
ysis assesses the efciency with which each muscle can
translate their respective generated force to spine stabil-
ity (Figure 7). Those muscles with higher values of the
efciency ratio have a greater normalized contribution
to spine stability for a given change in muscle force. In
contrast to the absolute impact of the various muscles to
spine stability, large efciency ratios were observed in the
multidus, quadratus lumborumand transverse abdomi-
nis, internal and external oblique, and iliocostalis lum-
borum produced. Relatively smaller values were ob-
served in the rectus abdominis, pars lumborum,
longissimus thoracis, and latissimus dorsi.
Discussion
Clearly, there is no single muscle that is superior at en-
hancing spine stability. In addition, the muscle manipu-
lation method described here has provided insight into
the potential neuromuscular control of lumbar spine sta-
bility. Results of this analysis indicate that muscles in the
trunk play several roles at once and that their roles de-
pend on the instantaneous demand placed on the spinal
column. Generally, those muscles that were antagonist to
the dominant moment of the task were most effective at
increasing stability. This nding supports the direction-
dependent cocontraction pattern that has been reported
in the more global muscles during different tasks.
14,29,30
The greatest reductions in stability were observed when
muscles that opposed the dominant destabilizing forces
were inactivated. For example, in a lateral bending task
such as the right side bridge, the dominant external force
at the spine is the ground reaction force acting at the
forearmthat forces the L4L5 joint in a left lateral bend.
The right abdominal muscles are activated not only to
oppose the left lateral bend moment to support the total-
body posture, but at the level of a single lumbar joint,
they potentially also protect against an instantaneous
instability resulting from an excessive rotation in lateral
bend. During an in vivo study by Cholewicki and
McGill,
31
the authors observed a temporary excessive
vertebral exion in a powerlifter who incurred an injury
while lifting. The authors hypothesized that a motor con-
trol error in a crucial back muscle may have been respon-
sible for the excessive exion instability. These tempo-
rary reductions may prevent those muscles, whose job
during a particular task is to oppose crucial destabilizing
forces and rotational instabilities, from controlling ver-
tebral motion. As the destabilizing forces on the spine
change through different postures, so do the muscles that
are able to oppose these forces.
1261 Torso Muscles and Rehabilitation Exercises

Kavcic et al
Across the various torso muscles, the mechanical ad-
vantage to provide stability to the lumbar spine varies
depending on the muscle. It appears as though, on aver-
age, the larger, more global muscles are better able to
alter spine stability than the smaller, intersegmental mus-
cles. This is most likely because of the larger force-
generating potential in these muscles and their ability to
generate higher levels of L4L5 compression, translating
to higher levels of spine stiffness. As well, their larger
moment arms enhance their ability to act as guy wires.
Interestingly, the increase in compressive loads on the
spine that result from muscular cocontraction has been
estimated to increase stability at a higher rate than the
additional compression. Specically, Granata and Mar-
ras
32
have estimated that stability is enhanced threefold
for a given increase in compression, whereas Grenier and
McGill
23
have computed the enhancement to be at least
twofold. It would appear that the qualitative assumption
that activating muscles that impose low compressive
loads as prime stabilizers is problematic when evaluating
a quantitative stability analysis. Among the more local
muscles, it is interesting to observe the minor ability of
the transverse abdominis to alter spine stability when
manipulated through its entire force- and stiffness-
generating abilities.
The mechanical advantage for the smaller, interseg-
mental muscles, particularly the multidus, and quadra-
tus lumborum, appears to come fromtheir efcient trans-
lation of generated force to spine stiffness and stability.
These results can potentially be explained by a phenom-
Figure 5. Increase in stability in-
dex resulting from activating a
muscle unilaterally to 100% MVC.
A, Results are shown for two
asymmetrical tasks. B, Decrease
in stability index resulting from
turning a muscle off bilaterally to
0% MVC. Results are shown for
two asymmetrical tasks. The
same pattern observed between
agonist and antagonist muscles
noted in Figure 4 is observed be-
tween right and left muscles dur-
ing asymmetrical tasks. R corre-
sponds to the muscle on the right
side; L corresponds to the mus-
cle on the left side.
1262 Spine

Volume 29

Number 11

2004
enon known as the follower load, described by Pat-
wardhan et al.
33,34
According to this theory, those mus-
cles that insert onto the spinal segments are better able to
translate their generated force along the compressive axis
of the spine or tangent to the curve of the lumbar spine.
It is important to note that previous work by Crisco and
Panjabi
10
reported the opposite, in that the more multi-
segmental muscles are more efcient at creating a critical
level of lumbar spine stiffness over the intersegmental
muscles. The discrepancy in ndings, however, is consis-
tent with the different models used. In their study, the
lumbar spine was modeled as a straight elastic column
with motion restrained to the frontal plane. With such a
linear model, a given level of activation in the multiseg-
mental muscles would impact many joints, whereas the
intersegmental muscles may only affect one or two joints,
as they described. In our analysis, the spine was modeled
with a natural lordotic curvature and motion existed in a
total of 18 df The ability of the intersegmental muscles to
follow the curvature of the spine and direct a large com-
ponent of force along the compressive axis is how the
efciency of these muscles dominated over the more mul-
tisegmental muscles.
When attempting to apply the results of this study to
clinical practice, consideration of the physiologic rele-
vance of this technique is necessary. In a human neuro-
muscular system, muscle synergies exist, where changes
in a given muscle activation level rarely occur in isolation
but rather are associated with changes in that of other
muscles. For example, Richardson and Jull
35
reported
Figure 6. A, Increase in L4L5 compression resulting from activating a muscle bilaterally to 100% MVC. B, Decrease in L4L5 compression
resulting from turning a muscle off bilaterally to 0% MVC. The larger, multisegmental muscles stabilize through their ability to generate
high levels of L4L5 compression, which is associated with increased levels of spine stiffness, together with their action as guy wires
enhancing the systems potential energy.
1263 Torso Muscles and Rehabilitation Exercises

Kavcic et al
that activation of the multidus is linked to that of the
transverse abdominis. In order to accurately assess the
stabilizing role of a given muscle and represent physio-
logic reality, synergistic patterns need to be considered.
Since this study examined the effect of changing a single
muscle, this could be considered a limitation in terms of
assessing synergies. Assessing the consequence of syner-
gies is much more difcult given the many roles each
component muscle plays, but it is our objective to exam-
ine this in the future. The benet of this analysis, how-
ever, is to address the clinical misconception that at any
given moment a single muscle can provide the necessary
stability to the lumbar spine. One nding observed in this
analysis was that no single muscle, when manipulated
from 0% to 100% MVC, created an unstable spine. It
does not seem reasonable then that any one muscle in
isolation has the capabilities to dramatically impact
spine stability, at least in the stability exercises tested
here, although we have found some low challenge tasks
where this is not the case. This study showed that as
loads are applied to the spine there is an integration of
the many different muscles in order to balance the stabil-
ity and moment demands, and these patterns change as
the spine loading patterns change.
One feature of this technique was that it was successful
at identifying the total contribution that each muscle can
make to stability, relative to the other torso muscles tested.
In this light, it is noticed that the smaller muscles have a
stabilizing role through their efcient generation of force,
however, as loads increase the need for the stronger global
muscles is required. One important note is that only the
contributions of the muscles force proles to stability were
assessed. Other potential roles that inuence stability such
as proprioceptive integration or passive-elastic link with
intra-abdominal pressure were not assessed.
Another benet of manipulating a muscle in isolation
is that the changes created in stability can be associated
with only the manipulated muscle. Allowing an optimi-
zation routine to balance the muscle force moments to
the external moments would have caused changes in the
force proles of various muscles, and associating these
changes to that observed in the stability index would be
a difcult task.
Many assumptions were made in this study as a result
of the biomechanical modeling procedure used. Assump-
tions made in the biomechanical models used in this
analysis have been documented previously,
11
and while
great attempts were made to achieve biodelity, this
highly complex analysis could not be performed without
them. Lastly, the conclusions of this work are limited to
the contrived stability exercises tested. One important
note is that in the abdominal curl exercise, the actual
torque is generated at the level of the midthoracic spine;
however, in this analysis, in order to assess the stabilizing
ability of the different muscles in the lumbar spine during
a exion task, the moment created at the thoracic level
was translated to the lumbar spine. Given that the study
participants were fully supported laying supine on the
oor, the individual muscle contributions to lumbar
spine stability during the type of abdominal curl per-
formed here remain unknown.
In terms of the practical application of the ndings in
this study pertaining to prevention and rehabilitation,
the clinical practice of isolated training of a specic mus-
cle or group of muscles in attempts to reduce the com-
pressive costs must be questioned. According to the re-
sults of this study, it appears justiable to train motor
patterns that involve the contribution of many of the
potentially important lumbar spine stabilizers. This seems
to be the case since although some of the highly regarded
Figure 7. RMS of stability curve normalized to RMS of muscle force curves. One mechanism explaining how the smaller, intersegmental
muscles stabilize could result from their ability to efciently translate their respective generated force to spine stability.
1264 Spine

Volume 29

Number 11

2004
local muscles can create stability very efciently, their abso-
lute contribution is not dominating and may not be suf-
cient during functional tasks. Focusing on a single muscle,
or only a few, appears to be misdirected clinical effort if the
goal is to ensure a stable spine.
Key Points
Using various assumptions and variations to a
biomechanical model, assessment of the stabilizing
role of different muscles was quantied for different
loading scenarios.
The role of each individual lumbar muscle
changes as the loads placed on the spine changes.
Consideration should be given to each potential
stabilizer when designing exercise programs in-
tended to increase spine stability.
References
1. McGill S. Low Back Disorders: Evidence-Based Prevention and Rehabilita-
tion. Champaign, IL: Human Kinetics, 2002:143.
2. Panjabi MM, Abumi K, Duranceau J, et al. Spinal stability and intersegmen-
tal muscle forces: a biomechanical model. Spine. 1989;14:194200.
3. Bergmark A. Stability of the lumbar spine: a study in mechanical engineering.
Acta Orthop Scand Suppl. 1989;60:154.
4. Richardson C, Jull G, Hodges P, et al. Therapeutic exercise for spinal seg-
mental stabilization. In: Lower Back Pain. London: Harcourt Brace, 1999.
5. Hodges PW, Richardson CA. Feedforward contraction of transverse abdo-
minis is not inuenced by the direction of arm movement. Exp Brain Res.
1997;114:362370.
6. Hodges PW, Richardson CA. Delayed postural contraction of transverse
abdominis in low back pain associated with movement of the lower limb.
J Spinal Disord. 1998;1:4656.
7. Hides JA, Stokes MJ, Saide M, et al. Evidence of lumbar multidus muscle
wasting ipsilateral to symptoms in patients with acute/subacute low back
pain. Spine 1994;19:165172.
8. Panjabi M, Abumi K, Duranceau J, et al. Spine stability and intersegmental
muscle forces: a biomechanical model. Spine. 1989;14:194200.
9. Wilke HJ, Wolf S, Claes LE, et al. Stability increase of the lumbar spine with
different muscle groups: a biomechanical In vitro study. Spine. 1995;20:192
198.
10. Crisco JJ, Panjabi M, The intersegmental and multisegmental muscles of the
lumbar spine: a biomechanical model comparing lateral stabilizing potential.
Spine 1991;16:793799.
11. Cholewicki J, McGill S. Mechanical stability of the In vivo lumbar spine:
implications for injury and chronic low back pain. Clin Biomech. 1996;11:
115.
12. Cholewicki J, VanVliet J IV. Relative contribution of trunk muscles to the
stability of the lumbar spine during isometric exertions. Clin Biomech. 2002;
17:99105.
13. Cholewicki J, Simons A, Radebold A. Effects of external trunk loads on
lumbar spine stability. J Biomech. 2000;33:13771385.
14. Gardner-Morse M, Stokes I. Trunk stiffness increases with steady-state ef-
fort. J Biomech. 2001;34:457463.
15. Gardner-Mores M, Stokes I. The effects of abdominal muscle coactivation on
lumbar spine stability. Spine. 1998;23:8692.
16. Granata K, Orishimo K. Response of trunk muscle coactivation to changes in
spinal stability. J Biomech. 2001;34:11171123.
17. McGill S, Norman R. Partitioning of the L4L5 dynamic moment into disc,
ligamentous and muscular components during lifting. Spine. 1986;11:666
677.
18. McGill S. A myoelectrically based dynamic three-dimensional model to pre-
dict loads on lumbar spine tissues during lateral bending. J Biomech. 1992;
25:395414.
19. Stokes I, Henry S, Single R. Surface EMGelectrodes do not accurately record
from lumbar multidus muscles. Clin Biomech. 2003;18:913.
20. Cholewicki J, McGill S. Relationship between muscle force and stiffness in
the whole mammalian muscle: a simulation study. J Biomech Eng. 1995;117:
339342.
21. White A, Panjabi M. Clinical Biomechanics of the Spine. Philadelphia: Lip-
pincott, 1978:79.
22. Ma SP, Zahalak GI. A distribution-moment model of energetics in skeletal
muscle. J Biomech. 1991;24:2135.
23. Grenier S, McGill S. Lumbar spine stability fromhollowing vs. bracing:
the transverse abdominis is no more important than any other muscle to
ensure lumbar stability. Submitted.
24. Howarth S, Allison A, Grenier S, et al. On the implications of interpreting the
stability index: a spine example. J Biomed. In press.
25. Winter DA. Biomechanics and Motor Control of Human Movement, 2nd ed.
Toronto: John Wiley and Sons, 1990:5657.
26. Delp SL, Loan JP, Hoy MG, et al. An interactive graphics-based model of the
lower extremity to study orthopaedic surgical procedures. IEEE Trans
Biomed Eng 1990;37:757759.
27. Arnold A, Salinas S, Asajawa D, et al. Accuracy of muscle moment arms
estimated from MRI-based musculoskeletal models of the lower extremity.
Comput Aided Surg 2000;5:108119.
28. Santaguida PL, McGill SM. The psoas major muscle: a three-dimensional
geometric study. J Biomech. 1995;28:339345.
29. Thomas JS, Lavender SA, Corcos DM, et al. Trunk kinematics and trunk
muscle activity during a rapidly applied load. J Electromyogr Kinesiol. 1998;
8:215225.
30. McGill S. Electromyographic activity of the abdominal and low back mus-
culature during generation of isometric and dynamic axial trunk torque:
implications for lumbar mechanics. J Orthop Res. 1991;9:91103.
31. Cholewicki J, McGill S. Lumbar posterior ligament involvement during ex-
tremely heavy lifts estimated from uoroscopic measurements. J Biomech.
1992;25:1728.
32. Granata KP, Marras WS. Cost-benet of muscle cocontraction in protecting
against spinal instability. Spine. 2000;25:13981404.
33. Patwardhan AG, Harvey R, GhanayemA, et al. Afollower load increases the
load-carrying capacity of the lumbar spine in compression. Spine. 1999;24:
10031009.
34. Patwardhan AG, Meade KP, Lee B. A frontal plane model of the lumbar
spine subjected to a follower load: implications for the role of muscles.
J Biomech Eng. 2001;123:212217.
35. Richardson C, Jull G. Muscle control-pain control: what exercises would
you prescribe? Man Ther. 1995;1:210.
1265 Torso Muscles and Rehabilitation Exercises

Kavcic et al