Cbkxd Biomechnics Vol. 11, No. 1. pp.

l-15,

1996
Copyright 0 1995 Elsevier Science Limited
Printed in Great Britain. All rights reserved
02684033196 $15.00 + 0.00

ELSEVIER

New Concepts and Hypotheses

Mechanical stability of the in viva lumbar

spine: implications
for injury and chronic low
back pain
J Cholewicki

PhD,

S M McGill Pm

Occupational
Biomechanics
and Safety Laboratories,
Department of Kinesiology,
Applied Health Sciences, University of Waterloo, Ontario, Canada

Faculty of

Summary-One
important mechanical function of the lumbar spine is to support the upper
body by transmitting
compressive
and shearing forces to the lower body during the
performance
of everyday activities. To enable the successful transmission
of these forces,
mechanical stability of the spinal system must be assured. The purpose of this study was to
develop a method and to quantify the mechanical
stability of the lumbar spine in viva
during various three-dimensional
dynamic tasks. A lumbar spine model, one that is
sensitive to the various ways that individuals utilize their muscles and ligaments, was used
to estimate the lumbar spine stability index three times per second throughout the duration
of each trial. Anatomically,
this model included a rigid pelvis, ribcage, five vertebrae,
90 muscle fascicles and lumped parameter
discs, ligaments
and facets. The method
consisted of three sub-models:
a cross-bridge
bond distribution-moment
muscle model for
estimating muscle force and stiffness from the electromyogram,
a rigid link segment body
model for estimating
external forces and moments acting on the lumbar vertebrae, and
an 18 degrees of freedom lumbar spine model for estimating
moments produced by
90 muscle fascicles and lumped passive tissues. Individual
muscle forces and their
associated stiffness estimated from the EMG-assisted optimization
algorithm,
along with
external forces were used for calculating the relative stability index of the lumbar spine for
three subjects. It appears that there is an ample stability safety margin during tasks that
demand a high muscular effort. However, lighter tasks present a potential hazard of spine
buckling,
especially
if some reduction
in passive joint stiffness is present. Several
hypotheses on the mechanism of injury associated with low loads and aetiology of chronic
back pain are presented in the context of lumbar spine stability.
Relevance-This
method allows one to analyse the overall stability of the multi-degree-offreedom in viva lumbar spine under a wide variety of dynamic, 3-D loads and postures.
Such a method is necessary to test new hypotheses implicating the dysfunction of the spine
stabilizing system (muscles, ligaments, and central nervous system) as a cause of certain
low back pain and injury cases. Under this type of analysis, a scenario is proposed to
explain injury that could occur during a light task - perhaps picking up a pencil from the
floor.
Key words:

Lumbar

C/in. Biomech.

spine, stability model

Vol. 11, No. 1, l-l

5, 1996

Introduction
One important mechanical function of the lumbar spine
is to support the upper body by transmitting compressive and shearing forces to the lower body during
the performance of everyday activities. However, the
isolated thoracolumbar spine buckles under comReceived: 8 November 1994; Accepted: 25 May 1995
Correspondence and reprint requests to: Jacek Cholewicki, Biomechanics Research Laboratory, Department of Orthopaedics, Yale
University School of Medicine, PO Box 208071, New Haven, CT
06520-8071, USA

pressive loads exceeding 20 N and the lumbar part of

the spine buckles under approximately 90 N*. But in
viva a spine may experience compressive loads ranging
from about 6000 N for more demanding everyday
tasks3 and up to 18000 N during competitive powerlifting4. Clearly, as an issue quite apart from injury
related to loading in excess of tissue tolerance limits,
mechanical stability of the spinal system must be of
concern at all levels of loading.
The crucial role of the trunks musculature to support
the spine similar to guy wires spanning a bending mast
is well documented5-.
In addition some have

C/in. Biomech.

Vol. 11, No. 1, 1996

suggested that intra-abdominal pressure contributes to

the overall stability of the spine6.*. These stabilizing
mechanisms are controlled by the neural system. It is
sensible to expect that in order to prevent buckling,
both the motor control system and the osteoligamentous spinal linkage will operate within the range of
mechanical stability, with some safety margin over and
above the critical load.
On the other hand there is evidence of cocontraction
of antagonistic trunk muscles during most daily
activities, which is energetically and mechanically
costiyc. Marc interestingly, such cocontractions
increase when people prepare for unexpected and
sudden loading such as an unknown weight dropping in
a hand-held bucket I. Any such cocontractions
increase joint Load and muscular energy requirement
for moment support. A plausible explanation is that a
certain level of muscle cocontraction may be necessary
to maintain the spine within a margin of safe, stable
equilibrium.
Indeed, new concepts were recently outlined that
link the mechanical instability of the lumbar spine to
the aetiology of clinical spine disorders. Panjabi,
postulated that the dysfunction of the spine stabilizing
system (muscles. ligaments, and central nervous
system) may lead to low back pain and injury.
Cholewicki hypothesized that the sudden need to
regain spine stability may result in excessive muscle
activity and tissue overload. Prolonged muscular
compensation to maintain the mechanical stability of
the spine may lead to chronic low back pain~. However, further progress in understanding the aetiology
of lumbar spine injury or pain in relation to its
mechanical stability requires an in viva quantification
of spine stability.
Previous research efforts have studied spine stability
using either optimization models to distribute muscle
forces.. or considered only the general effect of
muscle forcesi . These authors demonstrated that,
theoretically, regions of spine instability can occur
under certain conditions. However, current optimization models of the lumbar spine that minimize some
cost functions to obtain a unique equilibrium solution
and the resultant tissue loads are not suitable for in viva
stability analysis. The optimization approach inherently
lacks the biological sensitivity to various possible
muscle recruitment and cocontraction patterns. It
provides the same solution set of muscle forces for
everyone executing the same task, regardless if one
individual used his muscles in quite a different way.
The purpose of this study was twofold. Firstly we
developed a model overcoming the above difficulties.
Its features include anatomical detail (90 muscle
fascicles), together with a method that combines the
EMG-assisted approach with optimization to predict
muscle forces. A method for quantifying the mechanical
stability of the in viva lumbar spine is outlined.
Secondly we compared the mechanical stability of the
spine, in the context of clinical implications, in three
healthy subjects performing several three-dimensional,
dynamic tasks. It was hypothesized that the stability of

the spine is maintained by the neural system to a

constant level, regardless of various moment or joint
force demands and postures.
Methods

Three male subjects (age 24.3 years; SD, 2.6; height

1.77 m; SD, 0.03; body mass 82.2 kg; SD, 2.0) performed two trials for each of seven slow dynamic (to
minimize inertial interactions), three-dimensional tasks
(14 trials in total for each subject), with a load applied
to the hands via a dead weight routed through cables
and a pulley system. The tasks consisted of a twohanded lift off the floor (33.56 kg), trunk lateral
bending with the load in a right hand (20 kg), trunk
twist while holding the cable handle in front with outstretched arms (10 kg), one-handed push (10 kg), onehanded pull (10 kg), one-handed sweeping motion
from the floor up with a twist (5 kg), and upright
standing while the load held in both hands was
increased in three steps (0, 20, 40 and 60 kg total)
(Figure 1). Each task lasted 10 s with the exception of
upright standing, which was collected over a 12-s
period. In addition, all subjects performed four
maximum voluntary contraction tasks designed to
activate each muscle group maximally for normalization of the myoelectric signals20.
The subjects were videotaped with two synchronized
cameras, arranged approximately at right-angles to
each other. The joint centres of the subjects were
digitized 15 frames per second and low-pass filtered at
a cut-off frequency of 2 Hz. Residual analysis showed
that such movements were slow, smooth, and contain
virtually no signal above 2 Hz21. A calibration frame
(1.1 x 0.9 x 0.9 m) was used to calibrate a volume in
which the subjects were tested, yielding an average
error of 5 mm of marker reconstruction in any direction.
The joint centres of the body segments were digitized
directly from both video images, producing the threedimensional (3-D) coordinates of the actual joint
centres and not surface skin markers. These data were
input to a 3-D linked segment model (adapted from the
2-D version described in McGill and Norman2), from
which the upper body segment weights were obtained
together with the external forces acting on the ribcage
and in turn, on the lumbar spine (see flow chart Figure
2).

Fourteen channels of EMG were recorded using

bipolar, Ag-AgCl
surface, disposable electrodes,
placed with a centre-to-centre spacing of 3 cm over the
following muscles on each side of the body: rectus
abdominis (3 cm lateral to the umbilicus), external
oblique (approximately 15 cm lateral to the umbilicus),
internal oblique (approximately midway between the
anterior superior iliac spine and symphysis pubis, above
the inguinal ligament), latissimus dorsi (lateral to Tg
over the muscle belly), thoracic erector spinae (5 cm
lateral to Tg spinous process), lumbar erector spinae
(3 cm lateral to L3 spinous process) and multifidus
(2 cm lateral to L4-5 spinous processes). The signals
were A/D converted at a sample rate of 1200 Hz,

Cholewicki

and McGill:

Mechanical

stability

of the

in vivo

lumbar

spine

/
I, I
,

_*-**

I
,
I

c) SWEEP

d) STAND

e) TWIST

f) PULL

Figure 1. Dynamic tasks considered in this study: (a) two-handed

lift, (b) lateral trunk bend, (c)one-handed
load in the hands, (e) trunk twist, (f) one-handed pull, and (g) one-handed push.

3 (1

ANAIXSIS

18 DP

LUMBAR

SlfINE

MODEL

a
MUSCLE

standing

with increasing

Lumbar spine model formulation

1 DM

sweep motion,(d)

rectified, low-pass filtered, and normalized to the

maximum voluntary contraction values. Low-pass
filtering (second-order Butterworth filter) was performed using a cutoff frequency of 6 Hz to represent
muscle activation dynamics (calcium release and
diffusion process) with a time constant of about
25 ms23-25.Subsequently these EMG data served as an
input to the distribution-moment muscle contraction
dynamics mode126-28,40.

STABILITY INDEX
STABILTY

g) PUSH

ricrixl

DA3.A COLLECHON
Figure 2. Block diagram of the modelling procedure for the spine
stability analysis. This procedure consisted of three sub-models:
the
cross-bridge
bond distribution-moment
muscle model for estimating
muscle force and stiffness from EMG, the rigid link segment body
model for
-- estimating external forces and moments acting on the
and the 18 degrees of freedom lumbar spine model for
ribcage,
estimating moments produced by muscles and passive tissues described
here in detail. Estimated muscle forces were then adjusted with EMG
assisted optimization42
to achieve moment balance between the external
load acting on the ribcage and the muscle and passive tissue forces.
Adjusted muscle forces and stiffness together with the external forces
were needed for the stability analysis.

The anatomical representation and the calculation of

tissue loads were based extensively on the lumbar spine
model developed by McGill and Norman3 and refined
for 3-D analysis by McGi1129. However, substantial
extensions to this model were necessary, including full
anatomical development over the entire length of the
lumbar spine, to obtain viable input into the stability
analysis module. Original muscle cross-sectional areas
and attachments, obtained from cadaver dissections
and CT and MRI scans, were supplemented with data
from the recent work of Bogduk et a1.30.For details
see Appendix A.
Anatomical representation
The anatomical model consisted of a rigid pelvis and
sacrum, five lumbar vertebrae separated by a lumped
parameter disc and ligament equivalent for rotational
stiffness about the three axes, rigid ribcage and 90
muscle fascicles (Figure 3). Three axes of rotations
were assigned to each intervertebral joint for a total of

Clin. Biomech.

Vol. 11, No. 1, 1996

18 degrees of freedom (df) (6 joints x 3 df each).

Assuming a normal spine with no pre-existing translational laxity, joint translations were constrained to be
zero in the current version. Muscles were represented
with the centroid line approach; however, many of
them were forced to pass through several nodes
attached to the various vertebrae to account for the
curved lines of action. Each muscle consisted of an
active contractile part, a passive parallel elastic
element, and a passive non-linear tendon.
The amounts of lumbar lordosis (or lack of it) and
kinematics of the spine were derived from the relative
rotations between the ribcage and the pelvis measured
with an electromagnetic device 3SPACE ISOTRAK
(Polhemus Inc., Vermont, USA). In this way, the
model spine kinematics and range of motion were
calibrated to match each subjects spine. The
assumption was made that each intervertebral joint
contributes a constant proportion of the total rotation
angle between Si and Ti2 in all three directions: flexion/
extension, lateral bending, and axial twist (after White
and Panjabi31 and McGi112). Given the rotation angles
between the ribcage and pelvis, the new coordinates of
all lumbar vertebrae were obtained by rotating them in
the sequence about 2: X, and Y local (vertebrae) axes
(see Figure 3 for axes convention).
External load

The external forces acting on the ribcage and the points

of their application were obtained from a quasidynamic, three-dimensional
rigid linked-segment
model of a human body. Nineteen digitized joint
centres, reconstructed from a direct linear transformation algorithm contained in the 3-D video system
(Peak Performance, Colorado, USA), defined the 15
body segments (hands, forearms, arms, head, trunk,
pelvis, thighs, shanks, and feet). Segment lengths,
massesand centres of mass were taken from Winter32,
although parameters for the trunk and pelvis were
obtained from Zatsiorsky and Seluyanov. Two
additional markers indicated the direction of force
applied to the hands by means of a steel cable and a

6
F&we 3. Schematic diagram of the analysis of spine stability. Forces
acting onthe hands and torso (a) aretransmitted
to the ribcage and to the
lumbar spine and must be supported by torsional (passive tissue) and
linear springs (muscles) (b). The model (c)consists
of a rigid ribcage (0). 5
lumbar vertebrae Iwgments
1 to 5) and pelvis (6). Euler angles (rotations
about Z, X and Y axes) served as the generalized coordinates.

system of pulleys. The tasks were performed slowly and

smoothly, which permitted a static analysis of the
moments resulting from the body segment masses.The
hand force was measured directly with a transducer
mounted in series between the handle and a cable to
produce a quasidynamic analysis. Forces arising from
the mass of the hands, forearms, arms, head, and
trunk, their centre of mass coordinates, and the hand
forces were transformed into a local coordinate system
embedded in the pelvis. Together with the rotated
coordinates of the vertebral bodies, these forces were
summed to calculate moments about the disc centroids
at each intervertebral level. Subsequently passive tissue
and muscle forces were determined to balance the
moments calculated at every lumbar joint.
Passive tissue forces

The restorative moments contributed by passive tissues

(discs, ligaments, and other tissues) were calculated
first. Here the knowledge of intervertebral joint stiffness and a coupling of rotations throughout the entire
range of motion is essential. Unfortunately the data in
the literature are both sparse and scattered. A few 6 x 6
stiffness matrices for the isolated motion segments have
been reported, but pertain only to the stiffness about
the neutral joint position34,35.However, McGill et al.36
measured the passive trunk stiffness with respect to the
L4-5 joint, throughout the entire range of motion for
young males and females. These data incorporate the
passive contribution of fascia, skin, viscera, and other
tissues. The exponential shape of the positive joint
displacement/load relationship36,7, and a coupling
between the bending and twisting motion34 were
assumed. Given the absolute Euler angles as the
generalized coordinates, such a relationship for the jth
joint was expressed by:
M,j=uxj[ebo(~~--+)-

I] +K(I/I~--I+~~+,)

Myj=uyj[ebyl(~~--~+I)- l] +K(4j-~j+
Mzj= aZj[eb(~-~+f)-11

J
(14

Where:
Mij moment about ith axis of a jth joint,
a,b coefficients (they become negative for the negative angles),
K
coupling coefficient
In the case of trunk flexion, correction of the formula
was necessary. The data reported by McGill et a1.36
were collected while the subjects maintained bent
knees and a semi-seated position. According to
Andersson et a1.38,such a posture produces an average
of 28.4 of lumbar flexion when compared with a
standing posture. Subtraction of that angle in Equation
(1b) yielded a more realistic range of motion in the case
of trunk flexion:
Mzj=

-a,i{e[-bz~(%+B,

1,-28.41-c},

c = e( - 28.4bzJ
(13

Cholewicki

and McGill:

The coupling coefficient K was set between -8 N m/

rad34 and -46 N m/rad35 for all intervertebral joints.
The function shape coefficients (b Equation 1) were
taken from McGill et a1.36and the magnitude coefficients (a Equation 1) were adjusted to fit the
individual trunk flexibility of each subject in flexion and
lateral bending. This adjustment was accomplished by
measuring the lumbar flexion and lateral bending
angles and moments at the point where the trunk was
supported by passive tissues. Coefficients a were then
calculated to limit the range of motion at those angles
and to accommodate differences in subjects flexibility.
Coefficients for twisting and trunk extension remained
as the average values given by McGill et a1.36for all
subjects.
Muscle force and stiffness

The remaining moment (after accounting for passive

tissue contribution) necessary to balance the external
moment resulting from the hand-held load and upper
body weight, was partitioned between all 90 muscle
fascicles using EMG. Anatomically or functionally
similar muscle fascicles (muscle group) were assigned
the same EMG level. A previously documented EMGto-force processing approach3,29 has the disadvantage
that it does not yield muscle stiffness - a key
parameter in analysis of stability. Instead, a crossbridge bond distribution moment (DM) model was
employed here to obtain muscle force and stiffness
simultaneously. The DM is a mathematical approximation to Huxleys original two-state muscle contraction dynamics39. The additional advantage of this
model over Hills traditional equations is its ability to
mimic the muscle response to various dynamic stretchshortening input cycles. The most recent version of the
DM model was described by Ma and Zahalak28 and it
was enhanced here by adding a non-linear tendon and a
passive elastic component (see Cholewicki and
McGil14 for detailed description). The rectified filtered
and normalized EMG represented calcium release and
diffusion (activation dynamics), while the contraction
dynamics were modulated with the instantaneous
muscle length and velocity. The tendon slack length,
contractile tissue optimum (resting) length and the
maximum isometric muscle force of 35 N cm* constituted the input constants. This maximum stress was
selected only as the initial value, since all the muscle
forces were later adjusted with individual gains.
Several assumptions had to be made regarding the
neural activation of muscles not accessible for surface
EMG recording. Psoas and quadratus lumborum were
driven with the EMG signals of their synergists
(internal obliques and lower erector spinae respectively). Lafortune et a1.41 observed such synergy
patterns from many EMG electrode sites over the
lumbar spine during unconstrained three-dimensional
lifting tasks. These similarities could not be attributed
to electrical cross-talk. A more detailed discussion of
the above assumption can be found in Cholewicki et
al . l9. The activation patterns of small transversospinalis

Mechanical

stability

of the in vivo lumbar

spine

and rotatores muscles are unknown, but the small spine

stabilizing potential of these muscles could be simulated with. an increased passive joint stiffness5,7167.
Finally, each muscle force predicted from the EMG
was adjusted with an optimization algorithm to satisfy
the external moment requirements (within 10 N m
tolerance level). The objective of this method was to
balance all three moments acting about a given joint by
applying the least possible adjustment to the individual
muscle forces. This EMG-assisted optimization
(EMGAO) combines the advantage of both EMGassisted and optimization methods for estimating
muscle forces in an indeterminate biomechanical
model. It preserves physiologically observed muscle
recruitment patterns with an added advantage of the
balance of moment equationsi. The EMGAO procedure, discussed in detail in Cholewicki and McGi114*,
was expanded here to encompass 18 df and 90 muscle
fascicles. The individual muscle gains obtained from
this algorithm were applied to ajdust muscle forces as
well as their stiffness coefficients.
Stability analysis

The minimum potential energy method5,7,6,17,37was

carried to the first order (no post-buckling analysis)
considering three frames per second in every trial.
Conceptually, upon mathematical perturbation of the
spine in all directions, it is then explored whether it has
the ability to return to its original position. Mathematically a complete relative minimum of the potential
energy (V) of the system is a necessary and sufficient
condition to satisfy the mechanical stability criteria43. It
is equivalent to stating that the second variation of the
potential V must be positive definite44. Therefore the
determinant as well as the principal minors of the
Hessian matrix (second partial derivatives of the
potential V with respect to each of the generalized
coordinates Qi) must be positive (Equation 2). The
determinant D is called the stability determinant.

[ 1
a*v

D=det -

aQiaQj

a2v
a2v . . . a2v
aQ: aQlaQ2 am
a2v

= det

aQ2aQl

a2v

aQ:

*
...

a*v

aQnaQ1

>

OADij>O

a2v

aQ,z

(2)

Each data frame was frozen in time, while the

instantaneous spine geometry, external load, muscle
forces and their instantaneous stiffness constituted the
input to the analysis of stability (Figure 2). Upon
perturbation the potential of the system consisted of

Clin. Biomech. Vol. 11, No. 1, 1996

the elastic energy stored in the linear springs (U,)

(muscles and tendons), elastic energy stored in the
torsionat springs (Ur) (lumped intervertebral joint
discs, ligaments and other passive tissues) minus the
work performed on the external load (W):
v=C~,tC,--W

(3)

52
3

800-

600

5
3

Partiai derivatives of the potential V were calculated

(detailed calculations are presented in Appendix B)
and inserted to Equation 2.
Once the entire 18 x 18 Hessian matrix of the partial
derivatives of the systems potential energy was
compiled, it was diagonalized and its determinant
calculated. Elements of the diagonalized Hessian
matrix represent the curvature of the potential energy
surface in the direction of a given generalized coordinate; the larger the curvature, the more stable the
system is. If any value is less than or equal to zero,
instability of the spine is indicated. While the first mode
of buckling will possibly occur in the direction of the
lowest potential curvature, the stability determinant
describes better the depth of the entire 18dimensional surface in all directions. Therefore, the
18th root of the Hessians determinant served as a
relative (root average) stability index (SI). Figures 2
and 3 summarize the modelling procedure.
Resntts
The average maximum moments about the L4-s joint,
generated by the subjects, were 217 (SD, 28) Nm of
trunk extension in the lifting tasks, 114 (SD, 9) Nm of
lateral bend in the bending trial, and 54 (SD, 4) Nm of
twist in the twisting trial. Maximum Lhm5 joint compression forces occurred during the lifting trials and
averaged 3911 (SD, 211) N across the three subjects.
The intra-abdominal pressure (IAP), measured in one
subject, reached the highest values during the lifting
trials (189 mmHg). The lowest IAP peaks occurred
during the standing trials (44 mmHg).
The sweeping trials constituted a good test for the

lm!-

-+

-m- FLEXIEXT

LAT. BEND

!
I

2400

-2

Figure 5. Stability index (9) for all three subjects

were positive and high in magnitude.

Time (s)
Figure 4. Three-dimensional
moments (flexion/extension,
lateral bend,
axialtwist)
about the L4-s joint during a sweep trial of subject no. 1,
demonstrating
the loading challenge to thespine.

10

10

during

sweep

trials

model performance in complex 3-D tasks. There was a

mixture of all three moments acting on the lumbar
spine, which had to be balanced by the musculature
(Figure 4). Forces for the selected muscles and their
corresponding stiffness estimates in a sweep trial are
listed in Table 1. All major muscle groups were active,
indicating the high degree of muscle cocontraction. The
resultant Lde5 joint compression force varied between 2
and 2.5 kN, while the stability index (SI) characterized
the lumbar spine as a stable structure under such
conditions (Figure 5). Passive joint properties (disc and
ligaments) contributed little to the overall moment
requirements to balance the external load in all trials.
However, these passive tissues played a crucial role in
stabilizing the spine during tasks requiring little
demand on muscular forces. This phenomenon will be
discussed in greater detail later in the paper.
Sensitivity unalysis

Sensitivity analysis was performed on several parameters used in the model to determine their influence
on the stability index (SI). For example, normalizing
(or not) the adjustments according to the muscle size in
the EMGAO algorithm designed to balance the
moments had little effect on the SI (see Cholewicki and
McGi114 for the discussion of EMGAO) (Figure 6a).
Another sensitivity test was concerned with the site for
external load application. The hands holding the weight
were assumed to be rigidly attached to the ribcage in
this model, which is not the case in real life. When the

Muscle

Time (s)

Table 1. Estimated force and stiffness

second of the sweep trial of subject

--ff

R rect. abdominis
L rect. abdominis
R ext. oblique 1
L ext. oblique 1
R int. oblique 1
L int. oblique 1
R iliocost. lumb.
L iliocost. lumb.
R lat. dorsi L3
L lat. dorsi L3

of the selected muscles,

no.1 (Figure 4)
Force (NJ
74
39
66
6
18
48
109
67
55
81

Stiffness

for the 5th

IN cm
147
108
232
56
55
421
291
249
144
226

Cholewicki

SI

f-

SI (so1 2)

SI (2.7)

-++

SI (min)

and McGill:

7-Y

10

Time (s)

20 b

60 Lg

40 ke

6
Time (s)

10

12

Figure 6a. The effects of the stability index (SI) calculated as the 18th root
of the determinant
of the stability matrix, the passive joint stiffness
increase (2.7-fold), the alternate form of the objective function in the
EMGAO algorithm (Sol 2). and the alternate definition of the stability
index (SI min), where S1 was taken as the lowest value in the diagonalized
stability matrix, during a lifting trial for subject no. 1.6b. The effect of
passive joint stiffness increase (4.7 times) around a neutral spine position
and the effect of increasing some of the small intrinsic muscle activity
(multifidi and longissimus
thoracis/iliocostalis
lumborum pars
lumborum) from zero to 3% maximum.
The activation level of these
muscles, as indicated by their EMG, never exceeded 3%. However,
EMGAO algorithm would sometimes
reduce these forces to zero in order
to balance the moments. STAND2 trial for subject no. 1. Zero values of SI
indicate spine instability (SI<O).

simulated load was instead applied to the shoulder

region, the difference in the SI was smaller than 1%.
A more thorough analysis was performed on the
assumptions pertaining to the passive properties of
intervertebral joints. First, the effect of a lateral
bending and axial twisting motion coupling coefficient
(when varied from 0 to -46 in Equation la) on the SI
was determined to be less than lo%, even at the
extreme ranges of motion. Second, the experiments of
Crisco and Panjabi37 and Crisco et al.* on lateral
stability of the osteoligamentous spine were simulated.
The average buckling load of 97 (SD, 21) N obtained
from the 3 subjects corresponded well with the experimental value of 88 N (Crisco et al., 1992). However,
the objective of the lumped passive stiffness parameter
in this study was to encompass the effects of fascia,
viscera, skin, etc., in addition to the discs and ligaments. Crisco and Panjabi37 found that the exponential
joint stiffness model underestimated the joint stiffness
in a neutral position. In fact, they found it necessary to
increase the resting joint stiffness by an average of 2.7

Mechanical

stability

of the in vivo lumbar

spine

times to predict a reasonable critical load.

As in Criscos experiments, we presently found that
the stiffness of the local stabilizing system (muscles
attaching to the individual vertebrae and passive joint
stiffness5Y16)around the neutral joint position, was
insufficient to maintain spine stability during tasks
requiring minimal muscle forces. According to the
initial model output, spines of all the subjects appeared
unstable during some periods of the standing trial,
particularly under the heaviest load. The pushing tasks
were also characterized by low levels of stability, with
two subjects spines buckling during these trials.
The instances in which the subjects were unstable
were associated with the neutral postures and the near
zero muscle forces in the multifidi and the lumbar
erector spinae muscles (longissimus thoracis and iliocostalis lumborum pars lumborum) at the mid-lumbar
level (L&.
If the activity of these small muscles was
increased to at least l-3% MVC, the stability of the
whole system was restored (Figure 6b). Similar to
Crisco and Panjabis37 findings, the three- to five-fold
increase in the passive stiffness of the intervertebral
joints about their neutral position, assured stability of
the whole spine in all cases (Figure 6b). Perhaps this
may be an equivalent to the stiffening effect of the large
preload present in the in vivo lumbar spine46. However, the most important outcome from the sensitivity
analysis was that while the stability index was affected
by the passive joint stiffness, its relative magnitude was
always preserved (Figure 6).
Relative stability index (SZ)
The stability index (SI) exhibited two trends across all
subjects and trials. First, the stability increased with the
increased moment demand or the joint compression
force during the task (Figure 7). The lowest level of
stability was present when there was no demand on
high muscular forces, such as upright standing just prior
to the beginning of the lifting trial (Figure 7). In such
cases, the stability of the whole spinal system relied
predominantly on the passive stiffness of the inter-

,e
6
i

SI (Nmhdhad)

MOMENT

e
+

L4/LS

COMP.

(N)

(Nm)

4
$
2
y2

4
Time

10

(8)

Figure 7. Stability index (SI) followed a similar pattern in magnitude with

the L4-s joint compression
force and the trunk extensor moment (LIFT1
trial of subject no. 1).

Clin. Biomech.

Vol.

11, No.

1, 1996

vertebral joints. Second, in the asymmetrical tasks,

such as bending or twisting, the SI was also related to
posture (Figu-fe 8). The highest level of stability was
reached when the subjects were bent or twisted in such
a way that the external load placed on the lumbar spine
was acting thrtiugh the shortest moment arm. If the
load was-held in an outstretched arm, the sttibility level
was lower (Figure 8).
Discussion
The objective of this work was to develop a method for
quantifying the relative spine stability in viva. The need
for such a method arose from new hypotheses implicating the dysfunction of the spine stabilizing system
(mechanical stability) in the aetiology of some lumbar
spine disorders3,1. The method presented here can be
used to test these hypotheses and to study the clinical
issues related to the mechanical stability of the spine.
The stability index (SI) introduced in this study has
the units of stiffness per radian (Nm/rad/rad). It was
calculated as the 18th root of the product of 18 diagonal
elements in the diagonalized Hessian matrix. Each
element corresponds to the increase in stiffness arising
from rotating the spinal column in the direction of the
corresponding generalized coordinate. The SI then has
the physicai interpretation of a root average spine
stiffness slope in all directions. Since the spine is most
likely to buckle in the direction of the coordinate with
the lowest stiffness slope, it may be appropriate to also
define the SI as the lowest value in the diagonalized
Hessian matrix. The SI was calculated both ways;
however, as the relative measure, the SI defined either
way, follows a similar pattern (Figure 6a).
The most important finding of this study was that the
stability of thehuman lumbar spine seemed to increase
during the most demanding tasks as defined by the joint
compression force, and diminished during the periods
of low muscular activity. Thus the hypothesis that the
spine system maintains a constant stability safety
margin, regardless of the moment demands or
postures, was not supported by the results of this work.

Time (8)
Fire
8. Stability index(SI) and the lateral rotation of the ribcage in
reletion to the pelvis, demonstrating
that the SI is higher when the
externai toadvector
acts closer to the lumbar spine (BEND2 trial of
subject no. 1)

However, the control of spine stability suggested by the

present study is plausible, because the error of the
central nervous system and spine buckling at high loads
can be detrimental. On the other hand the stability
during very light tasks such as standing may not be as
critical, and perhaps is handled to a large extent by the
passive tissues to prevent the expenditure of energy.
While the large muscles spanning between the pelvis
and ribcage provide the bulk of stiffness to the spinal
column7,7, activity of the short, intrinsic muscles that
span only one or two joints (such as the multifidi and
the rotatores) was necessary to maintain stability of the
whole lumbar spine. Although the small transversospinalis and rotatores muscles were not modelled in this
study, their effect was simulated by increasing passive
joint stiffness. It was found that the spine will buckle if
the activity of multifidi and lumbar erector spinae is
zero, even when the forces in large muscles are substantial. Instability can be prevented by increasing the
activity (stiffness) of small muscles or by increasing
passive joint stiffness. It highlights the importance of
motor control to coordinate muscle recruitment
between the large musculature and the small intrinsic
spine muscles when handling small loads. These
observations are consistent with Bergmark5,16, who
lumped all passive joint properties and intrinsic muscles
into the local stabilizing system. He found that there
was an upper limit on the possible activation of the
large musculature at a given activation level of the local
stabilizing system, beyond which the spine buckles.
This work contributes to the explanation of spine
injury occurrence during activities requiring low
muscular forces (it is already quite easy to show how
injury occurs at high loads, where the load tolerance of
an individual tissue is exceeded). A hypothetical
relationship can be proposed, by which the injury risk
due to the loss of stability increases with the decreased
demand on muscular effort (Figure 9). For example,
this may explain how a person could work all day at a
demanding job and throw out his back picking up a

TASK DEMAND
(JOINT COMPRESSION)
Fiire
8. Hypothetical
model for injury risk to the spine due to tissue
failure and spine instability. While high loads can cause injury by tissue
disruption, instability at low loads may allow sufficient local joint
movement to overload or irritate soft tissues.

Cholewicki

and McGill:

pencil at the end of the day. The exact mechanism of

the injury is not known at this point, but it is hypothesized that the momentary loss of stability that would
lead to unexpected displacements, irritate nociceptors
in connective and/or soft tissues or nerve roots.
Conversely the sudden need to regain spine stability
may result in muscle spasm and overload of a single
tissue. If the motor control system faces the danger of
lumbar spine buckling, a plausible response would be
to activate a small intrinsic muscle(s) that crosses a
particularly unstable joint to counteract the large
displacements. In turn, this muscle is subjected to a
greater risk of tissue overload and injury. Larger
muscles spanning several joints are not suitable for such
action, as this would increase the load borne by the
inferior intervertebral joints, magnifying the effects of
buckling. We observed single vertebrae displacement,
confined only to rotation at L2-s, when we filmed a
powerlifter in vivo with videofluoroscope45. This study
confirmed that buckling behaviour can be limited to a
single lumbar level from inappropriate activation of
muscles.
The results of this study can also permit the formulation of several hypotheses on the aetiology of chronic
low back pain and support the theory proposed by
Panjabi13i4. Following the initial trauma, a decrease in
intervertebral joint stiffness or an increase in the
neutral zone may occur. The motor control system can
compensate to a certain degree with the additional
cocontractions of the intrinsic muscles, but this may
lead to muscle fatigue. Healthy people may be able to
rely on passive joint properties for spine stability during
standing or sitting, while clinically deficient patients
need to sustain prolonged static muscle contractions.
Perhaps the intrinsic spine musculature may itself be
deficient in providing sufficient stiffness to the spine
following nerve injury, muscle injury, or fatigue.
Another possibility is that people lacking motor control
skills commit errors during light activities and repeat
injuries to the same tissues. Clinicians need to explore
the effectiveness of motor control training as an adjunct
to the muscle strength improvement for reducing low
back pain episodes.
There are several limitations of the presented model.
The overall stability of the spine may be underestimated, because no effect of intra-abdominal
pressure was considered together with omission of the
transversus abdominis and small spinal rotator muscles.
On the other hand, the assumption that the spine is a
conservative system, overestimates the stability index
due to some losses of potential energy stored in viscoelastic structures after the perturbation. A set skeletal
geometry and muscle cross-sectional areas were used
for all subjects. It was not possible to record the activity
of intrinsic spine muscles such as psoas and quadratus
lumborum, although surface electrodes were strategically placed to act as surrogates. Our recent experience
with indwelling electrodes indicates that such choices
seem reasonably valid. No translations between the
intervertebral joints were allowed and the centres of
joint rotation did not migrate. We assumed a normal

Mechanical

stability

of the in vivo lumbar

spine

spine in our model, with no pre-existing translational

laxity. Although translations can be included in the
future, the addition of 18 degrees of freedom at present
would obscure the mathematical traceability of this first
attempt for quantifying the spine stability in vivo.
Finally, slow and smooth dynamic tasks (not ballistic)
justified application of static analysis of stability.
However, accelerations were considered in calculation
of the hand-held loads.
Model validity

It is not yet possible to validate directly human spine

models that predict muscle and other tissue forces. It
would be even more difficult to validate the model
quantifying spine stability in vivo. The present model
belongs to a certain category of models in science, for
which there are no tools for model validation47.
According to Lewandowski47, attributes of the present
work can be classified as a hard (built on the basis of
well-established natural laws), causal (discriminating
between cause and effect), and deterministic (as
opposed to probabilistic) modelling approach. The
modelled system exists in the real world (human spine),
but it is not possible to conduct direct experiments. For
this categorgy of models, Lewandowski47 proposes a
validation process consisting of component validation,
internal validity checks, sensitivity analysis and a
judgemental evaluation.
The component validation is based on the premise
that models built out of well validated submodels will
probably be valid47. This approach was applied to the
most critical variable in the calculation of spine stability
- muscle stiffness. Stiffness estimates obtained from the
sub-model of a muscle were consistent with the
experimental data in the literature4, which adds
confidence to the stability values.
Internal validity in hard models pertains to the
preservation of basic physical laws on which such a
model was built. For example, the advantage of using
Huxleys muscle model in the present work was that
each variable had a physical interpretation, allowing
the results to be inspected for consistency with the
initial assumptions4o. Energy balance and variational
mechanical principles were the main elements in the
stability analysis. Fulfilment of these laws was monitored in the model output by comparing the moment
balance with the first partial derivatives of the systems
potential, which should have been (and was) equivalent. Energy conservation surfaced in the output as a
symmetry in a Hessian matrix. A very significant check
for the internal validity of the whole modelling process
arose from the two independent estimates of moments
acting on the lumbar spine. The first was obtained from
the video image and a rigid linked body model and the
second came from the estimates of muscle forces and a
lumbar spine model (see Figure 2). It required an
average of 48% (SD, 27) adjustment to the individual
muscle forces to bring these two estimates to unanimity.
Considering that the dynamic, three-dimensional tasks
were studied, such a number indicates a good agree-

10

Chin. Biomech.

Vol. 11, No. 1, 1996

merit. In addition, since we initially assumed the maximum muscle stress of 35 N cm-2, the adjusted muscle
forces will result in less than 61 N cme2 of muscle
stress {mean plus one standard deviation adjustment)
even for maximally activated muscles. Such values fall
well within the physiological capacity of muscle force
generating potential quoted by various authors48.4y.
The anatomical representation of a lumbar spine and
its musculature also has a bearing on the model output. The subjects were selected to fit an average
morphology
assumed in this study. However, the
ultimate evidence of the models inaccuracy would be a
disagreement between the lumbar spine moments

healthy subjects. Clinical instability can be simulated in

the model as the decreased intervertebral joint stiffness
and necessary adaptations in the neuromuscular system
can be studied.
Acknowledgements

The helpful suggestions of Dr J Koorda regarding

structural stability analysis, and the constructive
comments of Drs K Norman and K Wells are greatly
appreciated. The financial assistance of the Natural
Science and Engineering Research Council, Canada
(NSEKC) is also acknowledged.

estimated hy the two independent procedures, discussed earlier.

Sensitivity analysis

was conducted

on the most

extensive assumptions regarding the passive properties

of intervertebral joints. Fortunately the motion segment stiffness had very little effect on the relative
values of SI, although the absolute values altered
(Figure 6). The relative spine stability index was also
unchanged by load placement or the moment balance
algorithm (Figure 6). Therefore conclusions regarding
spine stability. if based on the relative magnitude of SI,

References
1 Lucas DB, Bresler B. Stability of the Ligamentous Spine.
Report no. 40 from the Biomechanics Laboratory,
University of California, San Francisco, Berkeley, 1961
2 Crisco JJ, Panjabi MM, Yamamoto I, Oxland TR. Euler
stability of the human ligamentous lumbar spine: Part II
experiment. Clin Biomech 1992; 7: 27-32
3 McGill SM, Norman RW. Partitioning of the Lj-s
dynamic moment into disc, ligamentous and muscular
components during lifting. Spine 1986; 11: 666-78
4 Cholewicki J, McGill SM, Norman RW. Lumbar spine
load during the lifting of extremely heavy weights. Med

can be considered valid.

Finally the results obtained from this model, such as
the l-,4-j joint compression forces, can be considered
reasonable in view of similar results in literature.

SciSportsExerc
1991;23:1179-86
5 Bergmark A. Stability of the lumbar spine: a study in
mechanical engineering. Acta Orthop Stand 1989; 60

Although

6 Dietrich M, Kedzior K, Zagrejek T. Modelling of muscle

none of the previous

spine models was

directly validated. such a comparison may serve as a

,judgemental evaluation in the validation process
outlined by Lewandowski. Optimization models
generally underestimate joint compression force by
23-43% when compared with EMG-assisted models,
because they do not account for antagonist muscle
cocontraction. However, other EMG-assisted models
simulating similar tasks. result in similar spine compression forces as in the present study. For example,
Granata and Marras reported 2-3.4 kN L&r joint
compression

during

asymmetric

trunk

extensions

generating somewhat similar three-dimensional spine

moments as our sweep trial (Figure 4). In another
model. Mirka and Marras performed a stochastic
sensitivity analysis of various trunk muscle coactivation
patterns. They found that the lumbar spine compression force varied between 1.5 and 2.2 kN for
dynamic, 40 Nm of trunk extension moment. We
obtained between 2 and 2.5 kN of the L4m5 joint
compression force in the sweep trial, which compares
well with the values cited above.
We followed the validation process suggested by
Lewandowski for a category of models encompassing
the one presented here. There is no single validation
study for such models. However, the extensive testing
process which we followed, failed to demonstrate that
our approach to obtain a relative stability index is not
valid. This model can be used in the future to study
issues related to the mechanical stability of the spine.
The~most immediate question arising from this study
pertains

to differences

between

patients having motion

clinically

diagnosed
and

segment instability

[Suppl230]: l-54
action and stability of the human spine. In: Winters JM,
Woo SL-Y, eds. Multiple Muscle Systems: Biomechanics
and Movement Organization. Springer-Verlag, New
York, 1990; 451-60
7 Crisco JJ, Panjabi MM. The intersegmental and
multisegmental muscles of the lumbar spine: a
biomechanical model comparing lateral stabilizing
potential. Spine 1991; 16: 793-9
8 Tesh KM, Shaw-Dunn J, Evans JH. The abdominal
muscles and vertebral stability. Spine 1987; 12: 501-8
9 Marras WS, Mirka GA. Muscle activations during
asymmetric trunk angular accelerations. J Orthop Res
1990;8:824-32
10 Marras WS, Rangarajulu SL, Lavender SA. Trunk
loading and expectation. Ergonomics 1987; 30: 551-62
11 Lavender SA, Mirka GA, Schoenmarklin et al. The

effects of preview and task symmetry on trunk muscle

response to sudden loading. Hum Factors 1989; 31:
101-1s
12 Lavender SA. Preparatory response strategies seen prior
to sudden loading of the torso while in constrained
posture. In: Karwowski W, Yates JW, eds. Advances in
Industrial Ergonomics and Safety III. Taylor and Francis,
1991; 239-46
13 Panjabi MM. The stabilizing system of the spine. Part I.
Function, dysfunction, adaptation and enhancement.
1 Spinal Disord 1992; 5: 383-9
14 Panjabi MM. The stabilizing system of the spine. Part II.
Neutral zone and instability hypothesis. J Spinal Disord
1992;5: 390-7
1.5 Cholewicki J. Mechanical stability of the in vivo lumbar
spine. [PhD Dissertation] Dept. of Kinesiology,

University of Waterloo, Waterloo, Ontario, Canada, 1993

16 Bergmark A. Mechanical Stability of the Human Lumbar
Spine. [Doctoral Dissertation] Dept. of Solid Mechanics,
Lund University, Lund, Sweden, 1987
17 Crisco JJ, Panjabi MM. Postural biomechanical stability
and gross muscular architecture in the spine. In: Winters

Cholewicki

and McGill:

JM, Woo SL-Y, eds. Multiple Muscle Systems:

Biomechanics and Movement Organization.

Springer-Verlag, New York, 1990; 438-50

18 Shirazi-Ad1 A, Parnianpour M. Nonlinear response
analysis of the human ligamentous lumbar spine in
compression. On mechanisms affecting the postural
stability. Spine 1993; 18: 147-58
19 Cholewicki J, McGill SM, Norman RW. Comparison of
muscle forces and joint load from an optimization and
EMG assisted lumbar spine model: Towards development
of a hybrid approach. J Biomech 1995; 28: 321-31
20 McGill SM. Electromyographic activity of the abdominal
and low back musculature during the generation of
isometric and dynamic axial trunk torque: implications for
lumbar mechanics. J Orthop Res 1991; 9: 91-103
21 Wells RP, Winter DA. Assessment of signal and noise in
the kinematics of normal, pathological and sporting gaits.
Proc Canad Sot Biomech London, Ontario, Canada.
1980; 92-3
22 McGill SM, Norman RW. Statically and dynamically
determined low back moments during lifting. J Biomech
1985; 18: 877-85
23 Jiibsis FF, OConnor MJ. Calcium release and
reabsorption in the sartorius muscle of the toad. Biomech
Biophys Res Comm 1966; 25: 246-52
24 Bahler AS, Fales JT, Zierler KL. The active state of
mammalian skeletal muscle. J Gen Physioll967; 50:
2239-53
25 Zajac FE. Muscle and tendon: properties, models,
scaling, and application to biomechanics and motor
control. Crit Rev Biomed Eng 1989; 17: 359-411
26 Zahalak GI. A distribution-moment approximation for
kinetic theories of muscular contraction. Math Biosc 1981;
55: 89-114
27 Zahalak GI. A comparsion of the mechanical behaviour
of the cat soleus muscle with a distribution-moment
model. Biomech Eng 1986; 108: 131-40
28 Ma S-P, Zahalak GI. A distribution-moment model of
energetics in skeletal muscle. J Biomech 1991; 24: 21-35
29 McGill SM. A myoelectrically based dynamic
three-dimensional model to predict loads on lumbar spine
tissues during lateral bending. J Biomech 1992; 25:
395-414
30 Bogduk N, Macintosh JE, Pearcy MJ. A universal model
of the lumbar back muscles in the upright position. Spine
1992; 17: 897-913
31 White AA, Panjabi M. Clinical Biomechanics of the
Spine. Lippincott, Philadelphia, 1990; 106-12
32 Winter DA. Biomechanics of Human Movement. John
Wiley and Sons, New York, 1979; 151-2
33 Zatsiorsky V, Seluyanov V. The mass and inertia
characteristics of the main segments of the human body.
In: Matsui H, Kobayashi K, eds. Biomechanics VIII-B
Champaign, Human Kinetics Publishers, 1983; 1152-9
34 Panjabi MM, Brand RA, White AA. Three dimensional
flexibility and stiffness properties of human thoracic spine.

Mechanical

stability

of the in vivo lumbar

spine

11

J Biomech 1976; 9: 185-92

35 Patwardhan AG, Soni AH, Sullivan JA et al. Kinematic
analysis and simulation of vertebral motion under static
load - part II: simulation study. J Biomech Eng 1982;
104: 112-18
36 McGill SM, Seguin J, Bennett G. Passive stiffness of the
lumbar torso in flexion, extension, lateral bending and
axial rotation. Effect of belt wearing and breath holding.
Spine 1994; 19: 696-704
37 Crisco JJ, Panjabi MM. Euler stability of the human
ligamentous lumbar spine: Part I theory. Cfin Biomech
1992; 7: 19-26
38 Andersson GBJ, Murphy RW, Ortengren R, Nachemson
AL. The influence of backrest inclination and lumbar
support on lumbar lordosis. Spine 1979; 4: 52-8
39 Huxley AF. Muscle structure and theories of contraction.
In: Butler JAV, Katz B, eds. Progress in Biophysics and
Biophysical Chemistry. Pergamon Press, New York, The
Macmillan Company, 1957; 6-318
40 Cholewicki J, McGill SM. Relationship between muscle
force and stiffness in the whole mammalian muscle: a
simulation study. J Biomech Eng 1995; 117: 339-42
41 Lafortune D, Norman RW, McGill SM. Ensemble
average of linear envelope EMGs during lifting. Proc
Canad Sot Biomech, Ottawa, Canada, 1988; 92-3
42 Cholewicki J, McGill SM. EMG assistedoptimization:
a hybrid approach for estimating muscle forces in an
interdeterminate biomechanical model. J Biomech 1994;
27: 1287-9
43 Thompson JMT, Hunt GW. Elastic Instability
Phenomenon. John Wiley and Sons, New York, 1984
44 Langhaar HL. Energy Methods in Applied Mechanics.
John Wiley and Sons, New York, 1962
45 Cholewicki J, McGill SM. Lumbar posterior ligament
involvement during extremely heavy lifts estimated from
fluoroscopic measurements. J Biomech 1992; 25: 17-28
46 Janevic J, Ashton-Miller JA, Schultz AB. Large
compressive preloads decrease lumbar motion segment
flexibility. J Orthop Res 1991; 9: 228-36
47 Lewandowski A. Issues in model validation, Angewandte
Systemanalyse 3: 2-11. Reprinted in International Institute
For Applied Analysis 1982; RR-82-37
48 Weis-Fogh T, Alexander RM. The sustained power
output from striated muscle. In: Scale Effects in Animal
Locomotion. Academic Press, London, 1977; 511-25
49 Reid JG, Costigan PA. Trunk muscle balance and
muscular force. Spine 1987; 12: 783-6
50 McGill SM, Norman RW. Effects of an anatomically
detailed erector spinae model on L4-s disc compression
and shear. J Biomech 1987; 20: 591-600
51 Granata KP, Marras WS. An EMG-assisted model of
loads on the lumbar spine during asymmetric trunk
extensions. J Biomech 1993; 26: 1429-38
52 Mirka GA, Marras WS. A stochastic model of trunk
muscle coactivation during trunk bending. Spine 1993; 18:
1396-409

12

C/in. Biomech.

Vol. 11, No. 1, 1996

Appendix A: Anatomical
Parameters

model

of the muscles included in the model are listed in Table Al and coordinates

Tw
AI. Right body side muscles, their physiological
nodes (see Table A2 for the skeleton coordinates)

cross-sectional

Muscle
No.
___I___-.-1
2
3
4
5
6
7

9
10
?l
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45

YO
10.0
10.0
9.0
9.0
8.0
5.6
6.0
5.7
5.0
4.0
11 .o
16.8
0.7
1.8
12
1.5
1.3
2.0
1.0
1.0
1.0
1 .o
2.0
2.0
4.0
4.0
4.0
4.0
1.4
2.9
2.4
I.5
0.9
0.8
0.6
0.6
0.6
0.6
0.6
0.6
4.4
4.4
4.4
4.4
4.4

Rect. Abdominis
Ext. Oblique 1
Ext. Oblique 2
ht. Oblique 1
Int. Oblique 2
Pars Lumb. L;,
Pars Lumb. LR
Pars Lumb. Li
Pars Lumb. LZ
Pars Lumb. L,
Iiiocost. Lumb.
Long. Thor. P.
Long. Thor. L:
Long. Thor ii,
Long. Thor. L,
Long. Thor. L,
Long. Thor. L,
Quack. Lumb. P
Quadr. Lumb. L!
Quadr. Lumb. L2
Cluadr. Lumb. LJ
Ouadr. Lumb L4
Lat. Dorsi P
Lat. Dorsi L5
Lat. Dorsi L,
Lat. Dorsi L3
Lat. Dorsi L,
Lat. Dorsi L,
M&if.
P.LL,
M&if.
P.LJ
Moltif. P.L?
M&if.
P.L3
Multif. P.L,
M&if.
&5.L3
M&if.
L5.Ls
Multif. L5.L,
Multif. L4.LZ
M&if.
L4.L:
Multif. L3.1,
Multif. L2 TJ2
Psoas L5
Psoas L4
Psoas L3
Psoas L2
Psoas L!

Table A2. Coordinates

are zero)

defining

30.0
16.8
18.3
13.2
11.5
5.2
5.7
7,7
10.4
13.2
18.1
17.1
23.7
25.3
26.9
29.6
27.3
17.0
14.7
11.5
8.5
5.8
25.2
23.5
22.8
21.6
20.7
20.1
7.0
9.3
8.6
11,J
15.3
6.4
9.6
13.4
7.5
11.1
7.8
8.5
13.8
13.8
13.8
13.8
13.8

the skeletal

14.0
15.5
1.0
3.0
4.0
4.5
4.8
8.2
14.3
13.0
13.0
12.5
11.0
10.5
3.0
9.0
7.5
6.0
4.5
3.0
2.5
4.9
8.3
12.0
15.4

geometry

and muscle

of their origins, insertions

areas (no), resting

attachment

length (Xo), tendon

Origin

Insert

PEL (3)
PEL (4)
PEL (5)
PEL (6)
PEL (7)
PEL (9)
PEL (9)
PEL (9)
PEL (9)
PEL (9)
PEL (IO)
PEL (II)
L5 (4)
L4 (4)
L3 14)
L2 (4)
Ll (4)
PEL (12)
PEL (12)
PEL (12)
PEL (12)
PEL (12)
PEL (16)
PEL (15)
L4 (8)
L3 (8)
L2 (8)
Ll (7)
PEL (14)
PEL (14)
PEL (13)
PEL (13)
PEL (13)
L5 (6)
L5 (6)
L5 (6)
L4 (7)
L4 (7)
L3 (7)
L2 (7)
PEL (17)
PEL (17)
PEL (17)
PEL (17)
PEL (17)

RIB (3)
RIB (4)
RIB (5)
RIB (6)
RIB (7)
L5 (3)
L4 (3)
L3 (3)
L2 13)
Ll (3)
RIB (8)
RIB (9)
RIB (10)
RIB (11)
RIB (12)
RIB (13)
RIB (14)
RIB (15)
l-1 (5)
L2 (5)
L3 (5)
L4 (5)
RIB (17)
RIB (17)
RIB (17)
RIB (17)
RIB (17)
RIB (I 7)
L5 (5)
L4 (6)
L3 (6)
L2 (6)
Ll (6)
L3 (6)

and nodal points are presented in Table A2.

lengths

(Yo), points of origin,

insertion,

and

Nodal Points

L26)

54% RA length
64% RA length
-

on L4
on L4

h (10)
4 (IO)

Lq (11) L3 (IO) L* (10) L, (9)

i-4 (12)l-3(Ill Lz(11)L, (IO)
La (13) L3 (12) L* (12) L1 (I 1) RIB
L3 (12)L2(12)Ll (11) RIB
L2 (12) L, (II) RIB
L, (I 1) RIB

(23)
(23)
(23)
(23)

RIB
RIB
RIB
RIB
RIB

(22)
(21)
(20)
(19)
(18)

Ll 16)

L2P-3
Ll 6)
Ll (8)
RIB (16)
L5 (7)
L4 (9)
L3 (9)
L2(9)
Ll (8)

points

in a neutral

spine position

L,X(8)
L,X(9)
LSX(10)
L,X(ll)
LsX(12)
t,X(13)
L.X(l4)
L,X(l5)
L,X(16)

= 12
= 8.6
= 7.9
= 7.6
= 8
= 7.2
= 4.6
= 6
= 6.6

L,Y(8)
LsY(9)
L,Y(lO)
L,Y(ll)
L,Y(12)
L,Y(13)
LsY(14)
L,Y(l5)
L,Y(16)

L,,X(l)
4X(2)
L&(3)
L,&(4)
L,,X(5)
L,X(6)
L,X(7)
t&(8)
L,X(9)
L,X(lO)
L4X(11)
L&(12)
L4x(I3)
L4X(14)
LhX(15)
LdX(16)

=
=
=
=
=
=
=
=
=

La,Y(l) = 21.1
4Y(2) = 24.8
t,Y(3) = 23.4
tdY(4) = 22.4
LdY(5) = 23.8
L,Y(6) = 21.5
LdY(7) = 21.2
L4Y(8) = 22.4
L,Y(9) = 21.8
LdY(10) = 22.3
L,Y(ll)
= 22.3
L.,Y(12) = 22.3
t,Y(13) = 22.3
4Y(l4)
= 22.2
4Y(I5)
= 22.3
LY(I8)
= 22.6

(cm) (i.e. all relative

vertebral

rotations

PELVIS
PELX(11
PELX(2)
PELXI3)
PELX14)
PELX(5)
PELXIG)
PELX(7)
PELX(8)
PELX(9)
PELXIIO)
PELX(11)
PEt.X(?21
PELX(T3)
PELX(l4)
PELX(15)
PELX(16)
PELX117)
PELX(18)
fJELXll9)

= 10.4
= 9.4
= 18.4
=: 12.8
= 19
= 9
= 16
= 12.8
= 2.4
=- 1.4
-. 1.4
= 6
= 2.6
= 2
= 3.6
= 4.8
=z 15
= 1.3
= 5.4

PELY(I)
PELY(2)
PELY(3)
PELY(4)
PELY(5)
PELY(6)
PELY(7)
PEtY(8)
PELY(9)
PELY(10)
PELY(l1)
PELY(12)
PELY(l3)
PELY(l4)
PELY(I5)
PELY(16)
PELY(17)
PELY(l8)
PELY(l9)

=
=
=
=
=
=
=
=
=

6.8
17.4
5
18.6
5
21.5
16
18.6
17.8
= 16.6
= 16.5
= 21.4
= 18
= 13.8
= 19.2
= 21.5
= 5
= 16.6
= 21

9.4
10.6
7.6
4
4
5.8
8.6

L,Y(l)
L,Y(2)
L,Y(3)
L,Y(4)
L,Y(6)
LSY(6)
L,Y17)

17.4
21.1
20.4
20.4
19.9
19.1
18.8

PEtZ(1)
PELZ(2)
PEtZ(3)
PELZ(4)
PELZI5)
PELZiGj
PELZ(7)
PELZ(8)
PELZ(9)
PELZ(l0)
PELZ(11)
PELZ(I2)
PELZ(I3)
PELZ(l4)
PELZ(15)
PEtZ(16)
PELZ(l7)
PELZ(18)
PELZ(19)

=
=
=
=
=
=
=
=
=

7.6
0
3
13
0
12.5
12
13
6
= 6.8
= 3.3
= 9
= 3.6
= 1.5
= 3
= 6
= 8.2
= 0
= 6.2

LsZ(1)
L,Z(2)
L,Zi3)
L,Z(4)
L,Z(5)
L5Z(6)
L,Z(7)

0
0
5
0.2
0.5
1.5
2.3

hip joint
connect L5
rect. abd.
ext. obl. 1
ext. obl. 2
int. obl. 1
int. obl. 2
transv. abd.
pars. lumb.
iliocost. lumb.
long. thoracis
quad. lumb.
multifidus
01
multifidus
02
lat. dorsi sacr
lat. dorsi iliu.
psoas
lig. suprasp.
lumbod. fascia

L,X(l)
L,X(2)
L,X(31
La41
C&5)
bXf6)
L,XfJ)

=
=
=
=
=
=
=

=
=
=
=
=
=
=

=
=
=
=
=
=
=

= 19.2
= 20.2
= 20.9
= 20.4
= 21.4
= 22
= 20.8
= 21
= 21.2

L,Z(8) = 0
L,Z(9) = 0
L,Z(IO) = 0
L,Z(ll)
= 3.6
L,Z(l2) = 2
L,Z(I3) = 2.4
L,Z(l4) = 0
L,Z(15) = 0
L,Z(16) = 0

lig.
lig.
lig.
lig.
lig.
lig.
lig.
lig.
lig.

ant. long.
post. long.
flavum
intertransv.
caps.1 (lat.)
caps.2 (med.)
interspin.
interspin.
interspin.

hZ(ll

connect L5
connect L3
pars. lumb.
long. thoracis L,,
quad. lumb. L4
multifidus
(i.L.J
multifidus
(o.L,J
lat. dorsi L4
psoas
node PLl & PL2
node IL
node LT
node LT5
lig. ant. long.
lig. post. long.
lig. flavum

----J-4

L
connect PEL
connect L4
pars. lumb.
long. thoracis L5
multifidus
(i.L5)
multifidus
(o.Lg)
psoas

10.6
10.6
7.4
3.8
7.2
4.1
6.7
3.6
9.2
= 2.5
= 2.5
= 2.5
= 2.5
= 12.4
= 8.7
= 8

=
=
=
=
=
=
=
=
=

0
La4
0
L,Z(3)
4
LJ(4)
0.2
4Z(5)
4.4
t,Z(6)
0.6
L4Z(7)
1.5
t,Z(8)
0.2
L,Z(9)
2.3
L$Z(lO) = 5
L,Z(ll)
= 7.4
4Z(12) = 3.8
L,Z(13) = 1.5
LdZ(14) = 0
LaZ(15) = 0
4Z(16) = 0

Cholewicki

and McGill:

Mechanical

stability

of the in vivo lumbar

spine

13

Table A2 conri,

23.4
22.1
22
22
21.8
22.3
21.4
21.7
21.7
X
X

LJ(17)
L,Z(18)
LJ(19)
LJ(20)
42(21)
LJ(22)
LJ(23)
L,Z(24)
LJ(25)
L,Z(26)
L,Z(27)

10.6
9.8
6.9
3.2
6.2
4
6.3
3
8.8
= 2
= 2
= 2

L,V(l) = 24.8
LsVf2) -= 28.5
LsV(3) = 26.6
L,Vf4) = 25.2
LsY(5) = 26.8
LsV(6) = 24
LsV(7) = 24.3
LsY(8) = 25.2
LsV(9) = 25.3
LsV(l0) = 25.2
LsV(11) = 25.2
L,V(12) = 25.2

L,Zfl)
LsZ(2)
LsZf3)
LsZ(4)
LsZf5)
L,Z(6)
LsZ(7)
LsZ(8)
LsZ(9)
LsZ(l0)
LsZ(11)
LsZ(12)

=
=
=
=
=
=
=
=
=

9.8
8.7
5.9
2.6
5.2
3.2
6
2.4
7.7
= 1.4
= 1.4
= 1.4

L,V(l)
L,V(2)
L,V(3)
L,V(4)
L*Y(5)
LzV(6)
LzY(7)
LsV(8)
L,V(9)
L,V(lO)
LsV(11)
L,V(12)

L,Z(l)
L,Z(2)
L,Z(3)
LsZ(4)
LzZ(5)
LsZf6)
L,Z(7)
L,Z(8)
L*Z(9)
L,Z(lO)
L,Z(ll)
L,Z(12)

=
=
=
=
=
=
=
=
=

L.J(l7)
L&(18)
L&(19)
4x120)
LJ(21)
L,&(22)
LdX(23)
L,X(24)
LaX(25)
LX(26)
L&(27)

=
=
=
=
=
=
=
=
=
=
=

7.4
7.2
7.2
3.6
4.6
3.6
12.2
10.7
9
X
X

L4Y(17)
4Vf18)
L,rV(19)
4W20)
L4yf21)
L.,Y(22)
LV(23)
LV(24)
bV(25)
LV(26)
L4V(27)

=
=
=
=
=
=
=
=
=
=
=

=
=
=
=
=
=
=
=
=
=
=

3.6
2
1.6
0
0
0
0
0
0
X
X

lig. intertransv.
lig. caps.1 (lat.)
lig. caps.2 (med.)
lig. interspin.
lig. interspin.2&3
lig. supraspin.
ant. end plate
mid. end plate
post. end plate
talc. node for E02
talc. node for 102

0
0
3
0.2
3.8
0.5
1.5
0.2
2.1
= 7.6
= 3.9
= 1.5

connect L,r
connect Lz
pars. lumb.
long. thoracis Ls
quad. lumb. L3
multifidus
(i.Ls)
multifidus
(o.LJ
lat. dorsi Ls
psoas
node IL
node LT
nodeLT4and
LT5

0
0
2.7
0.2
3.8
0.5
1.5
0.2
2
= 7.8
= 4.1
= 1.5

connect Ls
connect L,
pars. lumb.
long. thoracis L2
quad. lumb. Lz
multifidus
(i.Ls)
multifidus
(o.LJ
lat. dorsi L2
psoas
node IL
node LT
node LT3, LT4 &
LT5

L3
L,X(l)
LsXf2)
LsX(3)
LaX(4)
LsX(5)
4X(6)
4X(7)
LsX(8)
LsX(9)
L,X(lO)
L,X(ll)
LsX(12)

=
=
=
=
=
=
=
=
=

L,X(l)
L,X(2)
L*X(3)
L,X(4)
LsX(5)
L2X(6)
L,X(7)
LsX(8)
L,X(9)
L*X(lO)
L,X(ll)
LsX(12)

=
=
=
=
=
=
=
=
=

=
=
=
=
=
=
=
=
=

28.5
32.2
29.8
27.8
30
26.9
27.4
27.8
29
= 27.9
= 27.9
= 27.9

Appendix B: Stability analysis

At any given frame, the potential of the spine system (V) is expressed as the sum
of the elastic energy stored in the linear springs (U,) (muscles and tendons),
elastic energy stored in the torsional springs (Ur) (lumped intervertebral joint
discs, ligaments and other passive tissues) minus the work performed on the
external load (W):

L1
L,X(l) = 8.7
L,X(2) = 7.2
L,Xf3) = 4.4
L,X(4) = 1.4
L,X(5) = 4
L,X(6) = 2.2
L,X(7) = 1.2
L,Xf8) = 6.6
L,X(9) = 0.2
L,X(lO) = 0.2
L,X(l 1) = 0.2

L,V(l)
L,V(2)
L,Y(3)
L,Y(4)
L,Yf5)
L,Vf6)
L,V(7)
L,V(8)
L,V(9)
L,V(lO)
L,Y(ll)

RIBX(1)
RIBX(2)
RIBX(3)
RIBX(4)
RIBX(5)
RIBX(6)
RIBX(7)
RIBX(8)
RIBX(9)
RIBX(l0)
RIBX(11)
RIBX(12)
RIBX(13)
RIBX(14)
RIBX(15)
RIBX(16)
RIBX(17)
RIBX(18)
RIBX(19)
RIBX(20)
RIBX(21)
RIBX(22)
RIBX(23)
RIBX(24)
RIBX(25)
RIBX(26)

= 7.2
= 7.5
= 19
= 6
= 12.5
= 15
= 19
R 1.6
= 2
= 2
= 2
= 2.5
= 2.5
= 2.5
= 3.5
= 1.8
= 9
= 1.2
= 2.4
= 3
= 3.6
= 3.6
= 0.2
= 18.2
= 11.5
= 4.6

RIBV(1)
RIBV(2)
RIBV(3)
RIBV(4)
RIBV(5)
RIBY(6)
RIBX(7)
RIBV(8)
RIBV(9)
RIBV(10)
RIBY(11)
RIBV(12)
RIBV(13)
RIBV(14)
RIBV(15)
RIBY(16)
RIBV(17)
RIBV(18)
RIBV(19)
RIBY(20)
RIBV(21)
RIBV(22)
RIBY(23)
RIBV(24)
RIBV(25)
RIBVf26)

s=

f [F,+K,(l,-l,,)]
m=l

= 0
= 0
= 2.6
= 0.2
= 3.6
= 0.5
= 0.2
= 1.9
= 8
= 4.3
= 1.5

connect L2
connect ribcage
pars. lumb.
long. thoracis L,
quad. lumb. L,
multifidus
(i.L,)
lat. dorsi L,
psoas
node IL
node LT
node LT2.3.4 & 5

= 35.5
= 68
= 35
= 30
= 31.5
= 29
= 38
= 39
= 44
= 53.5
= 58
= 62.5
= 66.7
= 68.5
= 35.5
= 34.2
= 47
= 34.5
= 31.7
= 29
= 26.4
= 24.2
= 34
= 37.6
= 37.2
= 34.8

RIBZ(1)
RIBZ(2)
RIBZ(3)
RIBZ(4)
RIBZ(5)
RIBZ(6)
RIBZ(7)
RIBZ(8)
RIBZ(9)
RIBZ(10)
RIBZ(11)
RIBZ(12)
RIBZ(13)
RIBZ(14)
RIBZ(15)
RIBZ(16)
RIBZ(l7)
RIBZ(18)
RIBZ(19)
RIBZ(20)
RIBZ(21)
RIBZ(22)
RIBZ(23)
RIBZ(24)
RIBZ(25)
RIBZf26)

= 0
= 0
= 7
= 12.5
= 10.5
= 7
= 0
= 8.4
= 5
= 2
= 2
= 2
= 2
= 2
= 7.2
= 0.5
= 12
= 6.5
= 6.5
= 6.5
= 6.5
= 6.5
= 1.5
= 0
= 0
= 0

connect L,
G
rect. abd.
ext. obl. 1
ext. obl. 2
int. obl. 1
int. obl. 2
iliocost. lumb.
long. thoracis
long. thoracis Ls
long. thoracis L4
long. thoracis L3
long. thoracis L2
long. thoracis L,
quad. lumb.
multifidus
(i)
lat. dorsi.
node LDI
node LD2
node LD3
node LD4
node LD (sacrum)
node LTl, 2,3,4,5
ant. diaphragm
mid. diaphragm
post. diaphragm

!!fk
da,

a2u, 90
-=
c 1K.~~+CF,+Kdr,~-r3]~
&tiaaj m=
J 1

(B4)

Since the partial derivatives are evaluated at the unperturbed point of

equilibrium, I,, -l,, = 0 and the Equations (B4) reduce to the following:

au, 90Fdl,,
-=
c maa,
aa, Ill=1

av au, au, aw
-=-.-.+---

ami au, au, au,

a2v azu, --~
a9,
a2w
~=aaiaa,+acliarj
aaiaaj

L,Z(l)
L,Z(2)
L,Z(3)
L,Z(4)
L,Zf5)
L,Zf6)
L,Z(7)
L,Z(8)
L,Zf9)
L,Z(lO)
L,Z(ll)

-RIBCAGE

v=u,+u,-w
Partial derivatives of the potential V were calculated separately for each component taking the Euler angles ai (3 rotation angles X 6 joints = 18 df) as the
generalized coordinates:

= 32.2
= 35.5
= 32.8
= 31
= 33
= 30.2
= 31
= 32.3
= 31
= 31
= 31

032)
If the muscle length is represented with a sum of n sections (when the muscle
passes through the nodal point), its potential energy derivatives consist of a
sum of its sections with some additional terms. Thus, if l,, = I,,, + I,,, + +
1omnand l,, = I,,, + l,,, + + lpmnthen

The energy stored in linear springs (U,) can be expressed as follows:

(B3)
where
F, = instantaneous muscle force (N),
K, = instantaneous muscle stiffness (N/m),
I,,,,,, Ir,,,, = original (frozen in a given frame) and perturbed muscle lengths
(4 and

W)

14

Ctin. Biomecb.

Vol. 11, No. 1, 1996

Since tbc length of a given muscle I, (dropping the muscle subscript m at this
pointJ is given by the vector sum of the length components in the X, Y and Z
axes direction,

(B7)
then

~=(l;,+l;,+l;,)-lz
(

l,,~+l,,~+lp,~
/ 1

w9

.I

where
h is a rotation matrix,
L is the vector of vertebral segment lengths taken between the adjacent joints,
X, Y, 2 are coordinates of the muscle attachment points in the reference
posture,
OX, OY, 02 are coordinates of the rotation centre (a joint) of a given
segment.
Partial derivatives of the elements of rotation matrices were easily programmed
on a computer by inserting the appropriate derivatives of the trigonometric
functions.
To obtain the elastic energy, which is stored in all of the torsional springs, we
need to integrate the Equation (1) with respect to the relative joint angles and
sum it over the 6 joints:

UTx=

i
j=O

Mxjl(~j-t$j+l)=,$o~

[ebw(h--cPl+)

Xl

and

UT,=

a21
l=
auiaaj

-(l;x+l;,+l;r)-32

i,lp*G+.i
dz ai
px--.E+
(\lhj

a1
1,,-g+l,~+l,,~

aat

byj

[ebj@+J+
2.1

-bzAej-ej+ 111

(B9)

W3)

The first partial derivatives of UT will have two terms belonging

adjacent intervertebral
joints:

to the two

= axj[ebd@f - 43+ l)- 11 + K($lj - ej+ 1)

Wj

(B6). (B7) and (B8) into (B4) yields

-~,(j-l)[&~-

I)(~~,I-~~)-l]-K(~j_l--j)

al

au,
~=F,(l;,+1;y+l;,)1~2

lb,

f?? [&(*J-$l+l)

III+ K(4j-4j+ IXll/j-+j+ 1)

M,d(B,--8,+,)=,jo
j=O

-au,
Substituting

i
j=o

-hyj($j-$j+

lpg~+lpz~
+(l;,+l;,+1;,)-2
%i
j )
-a21,, a1 al,,
px aaiaaj+xtf
z

a21
PY+P azPz
b a21,,
-I-PY
iJolidcljal.j ihi "aaiaclj>

M,j~(~j-~j+l)=

i
j=O

lpx,+l,~+lpz~
i

)
(B10)
au,
__=aZj[&(V4+1)-

aej

11+a,(j-l,[eb~-)(e~~~-e,)-

and

11
0314)

For the negative angles, coefficients a and b will appear with a minus sign
and the appropriate constants will be inserted in the case of flexion. Now, there
are six second partial derivatives of the UT possible for the general case:

azu, = -~,,j-l,~,,j- l)eb~-~)(+-+J)

Gja+j- I

--= au,

a$jWj+ 1
It remains to evaluate partial derivatives of muscle length components I,,, I,,,
I, in relation to all 18 rotation angles 01,. If the muscle originates on a skeletal
segment w and inserts onto the segment a (Figure 3). then its length vector

u, w=O,...6,

w>u

0312)

a2u,

- a ,b &,W-4,+4

Xl x.l

aw,

a~ja+j-l=a~ja*j+,=-

An identical equation format results if the UT formulation

of twist is
differentiated twice. Flexion/extension
has the same general format as (B15),
except K = 0 in this case.
The external work W performed by the load P is a dot product of the force
and displacement vectors:

Cholewicki

and McGill:

W= + *A%= P,(h,, - h,,) + Py(hpy- h,,) + ~,(h,, - h,,)

where h, and h, are the perturbed

Thus,

azw -F azh,,
-5&+F,g&+Fz~
auiauj

Mechanical

stability

of the in vivo lumbar

spine

15

Since the load P is always applied to the ribcage,

and the original points of force application.

W7)

The derivatives of the rotation matrix [IL] are the same in Equation (B12).
Because the global axes system is imbedded into the pelvis, the last term in
Equation (B18) vanishes upon the differentiation.
Once calculated, all partial
derivatives were inserted into the Hessian matrix in Equation (2).