Clinical Biomechanics 1992; 7: 27-32

Euler stability of the human ligamentous

lumbar spine. Part II: Experiment

JJ Crisco PhD1
MM Panjabi PhDl
I Yamamoto MD*
TR Oxland MASC’

‘Biomechanics Laboratory, Department of Orthopaedics and Rehabilitation, Yale University School of Medicine, New
Haven, Connecticut, USA; *Department of Orthopaedic Surgery, Hokkaido University School of Medicine, Sapporo,
Japan

Summary
The lateral backing and postbuckling behaviour of the intact and injured whole human
lumbar spine was experimentally studied using six fresh cadaveric specimens. The
ligamentous lumbar spine was loaded in axial compression and the lateral rotation of each
vertebra was recorded. At the point of the load application, the most superior vertebrae, the
specimens were constrained to move in the frontal plane since sagittal plane buckling will
not occur due to the lumbar lordosis. The average load required to buckle an intact whole
lumbar specimen was 88 N, and significantly decreased with injury. Once the spines had
buckled, the postbuckling behaviour was recorded. These results were compared to
theoretical predictions of a model (see Part I). The model was demonstrated to be in
excellent agreement with the experimental results.

Relevance
Demonstrated to behave as an Euler column, the ligamentous lumbar spine is unstable in
lateral bending under loads less than bodyweight. This reinforces the vital role of the
neuromuscular system and the dramatic effect its failure can have.

Key words: Euler stability, lumbar spine, model

Introduction in the inherent stability characteristics of the osteo-

To maintain upright posture is not a simple task, but it
is certainly a fundamental task of the neuromusculo-
skeletal system. Crucial to the success of this task is the
back. We may visualize the back as being composed of
two components: the osteoligamentous spine and the
neuromuscular system. Visualized as such, the osteo-
ligamentous spine appears as a column that supports
the large compressive loads’, and the muscles provide
additional stability to the column’. We are interested
Received: 19 November 1990
Accepted: 2.5 April 1991
Correspondence and reprint requests fo: Joseph J Crisco III, PhD,
Department of Orthopaedics and Rehabilitation, Yale University
School of Medicine. 333 Cedar Street. New Haven, Connecticut.
06510, USA
@ 1992 Butterworth-Heinemann Ltd
0268-0033/92/010027-06
ligamentous spine.
Lucas and Bresler3 hypothesized that the osteo-
ligamentous spine was similar to a long slender column
which, when sufficiently loaded with an axial com-
pressive force, would become unstable and buckle.
They first studied theoretically the mechanical stability,
as defined by Euler, of the cadaveric ligamentous
thoracolumbar spine, and calculated that the spine
could support no more than 20 N before becoming
unstable and buckling. They then validated their
calculations with experimentation on cadaveric speci-
mens. More recently Wilder4 studied the buckling
behaviour of single cadaveric lumbar spinal segments.
The whole lumbar spine, however, has not been
studied, and since the stability of a column is a discrete
function of both its length and its bending stiffness,
these previous studies cannot be used to study explicitly
other portions of the spine. Furthermore, to date, the
28 Clin. Biomech. 1992; 7: No 1

behaviour of the spine after buckling and the stability

of the injured spine has not been studied.
In Part I5 we examined theoretically the mechanical
or Euler stability of the osteoligamentous lumbar spine
in the frontal plane. We studied the buckling load and
postbuckling behaviour of the intact and injured spine.
The purpose of this study was to determine stability
characteristics experimentally using cadaveric speci-
mens, and to validate that model.

Methods
Six human lumbar cadaveric specimens were obtained
and stored freshly frozen (-20°C). Three specimens
were Si-L, and three were Si-Li. The average age of
the specimens was 55 years (range 50-62) and the
causes of death were cardiac or cerebral infarction. At
the time of testing each specimen (wrapped in a plastic
bag) was thawed in a warm water bath and dissected of
all muscular tissue. To fasten the specimen to the
experimental table and the loading jig, the sacrum and
the most superior vertebra were cast in a polyester
resin, from which two threaded studs extended. The
specimen was oriented such that the superior endplate
of L4 was horizontal. The weight of the superior casting

_-._ __-___.
b J’b I. _” vI--..-., - YV .-a... 11-v ‘y”-““v” I,..”

subjected only to its own weight when no axial

Figure 1. Posterior view of a specimen being tested
under the axial load of weighs (a) For this load the
specimen is stable; when perturbed it returned to this
upright position. (b) As the load is increased the
specimen became unstable and buckled. The linkage of
linear bearinos restrainina anterior-oosterior translation
can be seen a’t the top of e&h photograph.
compressive load was applied. This balancing required
two specific counterweights. The first counterweight
was fastened to the loading jig in such a manner that
the centre of mass of the mounting, counterweight, and
loading jig became the centre of the loading jig. The
second counterweight was attached to this point by
wires (Figure 1). The mass of the second counterweight
equalled the sum of the first counterweight, loading jig,
and mounting. The second counterweight travelled
freely in the frontal plane with the specimen as the
weight was suspended from linear bearings. The speci-
men was restrained from anterior-posterior translation
by the vertical shaft extending from the loading jig. The
shaft was incorporated into a linkage of linear bearings
that travelled on a similar horizontal shaft (Figure 1)
This system of counterweights and linear bearings
negated the loads of the mounting and loading jig as the
spine moved freely in the frontal plane. Details are
given elsewhere6.
The experimental protocol consisted of applying a
compressive axial load to the superior vertebra and
recording the relative vertebral motions.
The load was applied effectively to the centre of the
loading jig by two equal lateral loads (Figure 1). The
loads were dead weights, increased from zero, in
increments of 2 or 4 kg (20 or 40 N), up to the
maximum load that was considered safe for the
specimen. With each load increment the specimens
 Crisco et al.: Eular stability of spine: Experiment 29

were initially positioned in the upright posture and

released. The load step was maintained for 30 s-i
T
5-
before the vertebral displacements were recorded and
the load increased. 4-

To determine the relative vertebral motion we 3-

utilized methods well established in our laboratory,
2-
details of which are given elsewhere6. Briefly, we #_$ff$
used photo/radiographic vertebral markers and the l-
&y/c
principles of stereophoto/radiogrammetry to record the ?? "Url""""I
vertebral motions. From this data the complete three- 0 25 50 75 100 125 150 175
dimensional relative vertebral motion was determined. Axial load (N)
For the purposes of this study, only the relative Figure 2. The axial load-lateral rotation behaviour of an
vertebral rotations in the frontal plane were reported. intact (INT) lumbar spine specimen (Specimen No. 4) A,
Each spine was tested intact (INT), after injury to L,; +, L,; 0, L3; 0. L4; ??L5.
the L&i disc (DISC), and again after removal of the
L5 inferior facets (FAC). These injuries are described
in more detail in Part 15.

Results

All specimens demonstrated buckling behaviour as

shown in Figure lb. For a load less than the buckling
load (Figure la), the spine was stable: it remained
upright when perturbed. As the load was increased the
buckling load was reached and the spine buckled in the
lateral plane (Figure lb). In each specimen, buckling
occurred without a perturbation. The experimental Axial load (N)
findings will be presented in terms of the buckling loads Figure 3. The axial load-lateral rotation behaviour of an
and the postbuckling behaviour. injured (DISC) lumbar spine specimen (Specimen No. 4)
(See Figure 2 for key to legend).

Buckling loads
The buckling loads were determined for each specimen
and grouped with respect to the number of inter-
vertebral joints, four joints (L2-S1) in Table 1 and
five joints (L1-S,) in Table 2. For each specimen the
buckling load decreased with increasing severity of
injury. The average buckling load of the b-S1
specimen was greater than the L1-S1 specimens in the
INT and DISC, but was less in the FAC. Perhaps most
noteworthy was the average lateral buckling load for
Table 1. Buckling loads (N) of the L2-S, specimens
the whole intact ligamentous lumbar spine of 88 N.
Two factors influenced the determination of the
buckling loads. First was the error in the determination
of vertebral rotation. The error in this system was
determined to be approximately 0.606. Therefore, we
assumed that we could not accurately record the onset
of buckling until the rotations were greater than 1.2”, or
2 SD of the error. Second, the loads were applied in
discrete increments with weights. Therefore we defined
the approximation to the buckling load as the load at
which at least one vertebra rotated by more than 1.2”.
Defined as such, the buckling loads were upper bounds,
and rotations prior to buckling were defined to be zero.

Specimen INT DISC FAC

Postbuckling behaviour
1 118 98 39
2 98 78 20 The postbuckling behaviour was defined by the relation
3 78 39 39 between the axial load and the vertebral rotations.
Average 98 72 26 Since a discontinuity exists in this relation at the
buckling load (see Part 15), the data cannot be averaged
in the neighbourhood about the buckling load. Hence,
Table 2. Buckling loads (N) of the L1-S, specimens we illustrate the postbuckling behaviour by graphing
the results of a single specimen (No. 4). In Figure 2 the
Specimen INT DISC FAC rotations at each vertebral level of the INT specimen
are plotted as a function of the axial load. At 39 N the
4 78 39 20
5 - 78
specimen was stable. At 78 N the specimen was
39
6 98 59 59 unstable. As the load was increased the rotations
Average 88 59
increased. The largest rotations occurred at Ls, L4, and
39
L5, while the least rotation occurred at L,. After the
30 Clin. Biomech. 1992; 7: No 1

Table 4. Comparison of the experimental and theoretical

postbuckling rotations (degrees) at an axial load of 157 N

Ll L2 L3 L4 L5

01 e2 03 04 05

Experiment 1.9 (0.5) 3.1 (0.9) 3.3 (2.2) 3.3 (2.1) 3.6 (2.2)
Theory (Linear) 9.1 23.1 26.5 35.0 40.3
Theory (Exponential) 3.7 5.9 4.9 4.8 4.6

The rotations are given for each level of the intact (INT)
specimens and the models. For the experimental values, the
mean and (SD) are listed.

rotations at the maximum load applied. In Table 4 the

means and SD of the vertebral rotations of the whole
25 50 75 100 125 150 175 lumbar specimens are listed. Beneath these values are
Axial load (N)
the predictions of the two models. For each level it was
Figure 4. The axial load-lateral rotation behaviour of an readily apparent that the linear model predicted
injured (FAC) lumbar spine specimen (Specimen No. 4) rotations that were excessive, whereas the values
(See Figure 2 for key to legend).
predicted by the exponential model were within several
degrees of the average experimental values. For
Ls-Sr disc injury (DISC), the rotation at each level
example: the average L5 rotation was 3.6” at 157 N.
increased relative to the intact (Figure 3). The rotations
The linear model predicted a value of 40.3”. In
of the Lr, L, Ls, and L4 were similar, while the
contrast, the exponential model predicted a value of
rotation of L5 was largest. With further injury (FAC),
4.6”.
the rotations increased further (Figure 4). Although the
Similar results were found for the injured specimens
magnitude of the rotations increased with injury, the
at the maximum applied loads of 98 N for DISC and
behaviour was similar in each of these three graphs.
78 N for FAC (Tables 5 and 6).
In each of these Figures, the data points are
In Figure 5 we summarize the comparison of the
connected by lines for ease of visualization. The solid
experimental (specimen No. 4) and theoretical results.
lines are a fair interpolation of the data once the
Here we have plotted the vertebral rotation of the
specimen has buckled. However, the dotted lines
intact L5 as a function of the axial load. The experi-
indicate the existence of a discontinuity. A dotted line
mental data shows that the buckling loads is between
is used to indicate that the position of the discontinuity,
39 and 78 N. The buckling load predicted by the linear
i.e. the buckling load, cannot be determined exactly,
model lies in this range, in good agreement with the
only bracketed by the discrete load values.
experimental results. The buckling load predicted by
the exponential model is significantly less than this
Comparison of theoretical and experimental results range. For both these curves the values were plotted at
In Table 3 we have listed the theoretical buckling loads
calculated in Part I5 and the average experimental Table 5. Comparison of the experimental and theoretical
values of the whole lumbar spine. For the intact and the postbuckling rotations (degrees) at an axial load of 98 N
injured specimens the buckling loads were consistently
greater than those predicted by either the linear or Ll L2 L3 L4 L5

exponential models. The predictive ability of the linear e1 e2 e3 04 e5

model improved as the severity of injury increased;

Experiment 1.4 (1.1) 3.7 (2.3) 3.7 (1.8) 3.4 (1.6) 6.5 (3.0)
however, the exponential model predicted values that Theory (Linear) 6.1 15.1 16.8 22.4 69.4
were approximatley two orders of magnitude leas. Theory (Exponential) 3.4 5.3 4.5 4.4 9.4
The ability of the models to predict the postbuckling
The rotations are given for each level of the injured (DISC)
behaviour was evaluated by comparing the vertebral specimens and model.

Table 3. Comparison of the experimental and theoretical Table 6. Comparison of the experimental and theoretical
buckling loads (N) of the intact and injured ligamentous postbuckling rotations (degrees) at an axial load of 78 N
lumbar spine (L,-S1)

L7 L2 L3 L4 L5
Injury Experiment Theory
81 02 03 04 e5

Linear Exponential Experiment 1.7 (1.5) 3.5 (2.2) 3.4 (1.6) 3.7 (2.0) 7.2 (4.2)
Theory (Linear) 5.6 13.6 14.8 19.2 65.1
INT 88 66.8 11.0 Theory (Exponential) 3.1 5.0 4.2 4.1 10.1
DISC 59 44.5 0.2
FAC 39 41.8 0.2 The rotations are given for each level of the injured (FAC)
specimens and model.

9
3 injured vertebra.
8 8
b When the specimens did buckle, they buckled with
7
g c”
5
4
3
2 specimen. In general, the magnitude of the vertebral
1 rotation was larger the more inferior the vertebra
5
Axial load (N)
Figure 5. The buckling behaviour predicted by the linear,
A, and exponential, 0, models are plotted with the
experimental, W, values (Specimen No. 4) for direct
comparison. These data are for the intact L5 level.
10 N intervals, and hence these plotted data also
bracket the actual predicted buckling load.
Although the linear model was a good approximation
to the buckling load, the model predicted excessive
postbuckling rotation of Ls. The exponential model, on
the contrary, predicted postbuckling behaviour that is
in excellent agreement with the experimental data.
The human lumbar spine is lordotic in upright posture.
As is readily demonstrated upon studying lumbar
cadaveric specimens, this midsagittal curvature is
inherent to the ligamentous spine, which consists of
osseous, ligamentous, and cartilaginous tissue struc-
tures. If upright ligamentous cadaveric specimens are
subjected to an axial load that remains vertical, then
the lordosis will increase as the magnitude of the load
increases. This increase in the lordosis will continue
until there is a material failure of one or more tissue
structures. Such continuously increasing displacement
is characteristic of a deforming structure but not of the
instability phenomenon of bifurcation in an Euler
column.
In an Euler column, the column remains at a single
equilibrium position until the buckling load, which is
significantly less than the material failure load, is
reached. Under increasing axial loads the ligamentous
spine remains in an upright posture only relative to the
frontal plane. Thus, we restrained the specimens to
remain in the frontal plane, as Lucas and Bresler3 had
done with thoracolumbar specimens. Unlike Lucas and
Bresler, who utilized strings between the specimen and
a midsagittal vertical rod, we utilized a linkage of linear
bearings.
The buckling loads of the L2-Si and Li-Si
specimens were averaged separately because, as shown
in Part I”, the buckling phenomenon is dependent upon
the number of vertebral bodies. We found that the
intact L2-Si specimens were more stable - they
buckled at higher loads - than the Li-Si specimens.
The stability of each specimen decreased with in-
creasing severity of injury. This decrease in stability
Crisco et al.: Eular stability of spine: Experiment 31
was associated with a decrease in the stiffness of the
each vertebra displaced to the same side, as shown in
Figure lb. In none of the experiments was a complex
buckling shape, an ‘S’ shape for example, recorded.
The postbuckling behaviour was also similar for each
because of the increased moment arm as the specimen
bends and the load remains vertical. With injury at
Ls-Si, the rotation at each of the other intact levels
increased, but the rotation at L5-S1 increased substan-
tially more.
Previously, Lucas and Bresler3 demonstrated that
the buckling load of the intact cadaveric thoracolumbar
spine (devoid of ribs) can be predicted with a
continuous, homogeneous Euler column model. In this
two-part study, we have shown that a discrete, non-
homogeneous Euler column qualitatively predicts not
only the buckling load of the intact cadaveric lumbar
but also the buckling load of the injured spine and the
postbuckling behaviours.
The models predicted that the shorter spines are
more stable, in accordance with Euler’s theory as the
buckling load is inversely proportional to the height of
the column. The models predicted a decrease in
stability with injury since the buckling load of an Euler
column is proportional to its stiffness. The models
predicted that the components of the eigenvector
associated with the buckling load were all of the same
sign, i.e. the simplest buckling shape. The models also
predicted a postbuckling non-linear axial load-rotation
behaviour.
Although each model predicted the observed be-
haviour qualitatively, there were significant differences
between the abilities of the linear and exponential
models to predict the observed experimental behaviour
quantitatively. The difference between the linear and
exponential models, formulated in Part I’, was the
simulation of the elastic behaviour of the intervertebral
joint. We believe that the discrepancies between each
model and between the experimental data, summarized
in Figure 5, can be attributed to the simulation of the
intervertebral joints.
The elastic behaviour of the joints was determined by
applying a pure lateral bending moment and deter-
mining the relative vertebral rotations using well-
established techniques in our laboratory6. That data,
referred to henceforth as the flexibility data, was then
fitted with a linear and an exponential function (Figure
4 in Part I’). The stiffness of the joint is the inverse of
the slope of these functions, which is constant with load
for the linear function and increases with load for the
exponential function.
The buckling load is associated with the stiffness at
zero load; therefore, as expected from the flexibility
data, the buckling load of the linear model is greater
than that of the exponential model. However, even
though the exponential function fits the flexibility data
32 Clin. Biomech. 1992; 7: No 1
extremely well, the buckling load predicted with the
linear model was significantly better (Figure 5). We
believe this disparity resulted from the preconditioning
of the specimens prior to recording the flexibility data,
whereas prior to recording the buckling data, the
specimens were not preconditioned. This lack of
preconditioning would result in the exponential model
underestimating the stiffness about the neutral
position. Above a zero load, the stiffness of the
exponential function quickly becomes greater than the
linear stiffness value; therefore, the postbuckling
rotations of the exponential model are less than those
of the linear model.
Cyclic loading and unloadings or preconditioning
prior to recording data, although not well understood,
is a common preliminary procedure for ensuring
reproducible elastic behaviour in soft tissue. Precondi-
tioning is known to shift the entire load-displacement
curve along the displacement axis, thereby decreasing
the initial stiffness (increasing the ‘toe’ region’).
Beyond the toe region the elastic stiffness is not
influenced by preconditioning. This also appears to
be how the vertebral joints were affected in this
experiment.
We suggest that the non-preconditioned specimen
was stiffer than the preconditioned specimen about the
upright position, and that preconditioning does not
affect the stiffness of the spine to either side of the
upright posture. We believe that this explains why the
linear model best predicts the buckling load and the
exponential model best predicts the postbuckling
behaviour (Figure 5). It also appears that a biphasic
linear approximation of the joint stiffness, in which the
stiffness above 2 N m was approximated by a second
linear stiffness value, would give good postbuckling
predictions. We might further hypothesize that had the
specimens been preconditioned the buckling load
would have been substantially less than the 88 N
recorded here. The specimens were not preconditioned
prior to this buckling experiment because we did not
foresee this phenomenon.
A similar phenomenon was reported by Wilder4. He
experimentally determined the buckling load of single
preconditioned vertebral joints to be approximately
500 N. Based upon this buckling load, he calculated
an intervertebral stiffness of 99.56 N mm-‘, which
was substantially less than the stiffness value he
determined experimentally on non-conditioned joints
(2100 N mm-‘).
Lucas and Bresler3, in an internal report, predicted a
buckling load of 20.5 N and measured experimentally a
value of 19 N. These results are surprisingly in close
agreement considering that they modelled the thora-
columbar spine as a continuous column (no rigid
bodies) with a stiffness that was the mean of all levels.
In Part I5 we showed that the spine optimizes the
increase in stiffness from the cervical to lumbar region
to support a greater load than can be supported by a
column, continuous or discrete, with a mean stiffness.
There are believed to be no other studies that have
examined the buckling loads of the spine, and this is
believed to be the first that has studied the postbuckling
behaviour of the multiple-level spine.
Low back pain has been partially attributed to
clinical spinal instability. However, clinical spinal
instability is a complex pathology with no generally
accepted definition ‘-I2 , Although many definitions of
spinal instability include mechanical parameters, we
make no attempt at this time to correlate definitions of
clinical spinal instability with the Euler stability
presented here.
References
1 Nachemson A. The load on lumbar discs in different
positions of the body. Clin Orthop 1966; 45: 107-22
2 Crisco JJ, Panjabi MM. Postural biomechanical stability
and gross muscular architecture in the spine. In: Winters
JM, Woo SL-Y, ed. Multiple Muscle Systems.
Biomechanics and Movement Organization. New York:
Springer-Verlag, 1990: 438-50
3 Lucas DB, Bresler B. Stability ofthe Ligamentous Spine.
Technical Report esr. 11 No. 40, Biomechanics
Laboratory, University of California San Francisco, The
Laboratory, 1961
4 Wilder D. On loading of the Human Lumbar
Intervertebral Motion Segment. [PhD Dissertation]
Department of Civil and Mechanical Engineering,
University of Vermont, 1985
5 Crisco JJ, Panjabi MM. Euler stability of the human
ligamentous lumbar spine. Part I: Theory. Clin Biomech
1992; 7: 19-26
6 Crisco JJ. The Biomechanical Stability of the Human
Lumbar Spine: Experimental and Theoretical
Investigations. [PhD Dissertation] Department of
Engineering and Applied Science, Yale University, 1989
7 Fung YC. Biomechanics: Mechanical Properties of Living
Tissue. New York: Springer-Verlag, 1981: 223.5
8 Holdsworth F. Fractures, dislocations, and
fracture-dislocations of the spine. J Bone Joint Surg 1970;
52-A: 1534-51
9 Whiteside TE. Traumatic kyphosis of the thoracolumbar
spine. Clin Orthop 1977; 128: 78-92
10 Barford NC. Experimental Measurements: Precision,
Error and Truth. London: Addison-Wesley Publishing,
1967
11 Kirkaldy-Willis WH, Farfan HF. Instability of the lumbar
spine. Clin Orthop 1982; 165: 110-23
12 White AA, Panjabi MM. The clinical biomechanics of the
occipitoatlantoaxial complex. Orthop Clin North Am
1978; 9: 867-79