COPYRIGHT 2003

BY

THE JOURNAL

OF

BONE

AND JOINT

SURGERY, INCORPORATED

A Dynamic Study of
Thoracolumbar Burst Fractures
BY RUTH K. WILCOX, PHD, THOMAS O. BOERGER, FRCS, DAVID J. ALLEN, FRCS, DAVID C. BARTON, PHD,
DAVID LIMB, BSC, FRCSED(ORTH), ROBERT A. DICKSON, DSC, AND RICHARD M. HALL, PHD
Investigation performed at the School of Mechanical Engineering, University of Leeds,
and Musculo-Skeletal Services, St. Jamess Hospital, Leeds, United Kingdom

Background: The degree of canal stenosis following a thoracolumbar burst fracture is sometimes used as an indication for decompressive surgery. This study was performed to test the hypothesis that the final resting positions of the
bone fragments seen on computed tomography imaging are not representative of the dynamic canal occlusion and
associated neurological damage that occurs during the fracture event.
Methods: A drop-weight method was used to create burst fractures in bovine spinal segments devoid of a spinal
cord. During impact, dynamic measurements were made with use of transducers to measure pressure in a synthetic
spinal cord material, and a high-speed video camera filmed the inside of the spinal canal. A corresponding finite element model was created to determine the effect of the spinal cord on the dynamics of the bone fragment.
Results: The high-speed video clearly showed the fragments of bone being projected from the vertebral body into the
spinal canal before being recoiled, by the action of the posterior longitudinal ligament and intervertebral disc attachments, to their final resting position. The pressure measurements in the synthetic spinal cord showed a peak in canal pressure during impact. There was poor concordance between the extent of postimpact occlusion of the canal as
seen on the computed tomography images and the maximum amount of occlusion that occurred at the moment of impact. The finite element model showed that the presence of the cord would reduce the maximum dynamic level of canal occlusion at high fragment velocities. The cord would also provide an additional mechanism by which the
fragment would be recoiled back toward the vertebral body.
Conclusions: A burst fracture is a dynamic event, with the maximum canal occlusion and maximum cord compression occurring at the moment of impact. These transient occurrences are poorly related to the final level of occlusion
as demonstrated on computed tomography scans.
Clinical Relevance: In a thoracolumbar burst fracture, the final position of the fragments, as seen on computed tomography images at presentation, probably does not represent the maximum level of canal occlusion or peak cord pressure and therefore does not represent the probable damage to the cord tissue that occurred at the moment of impact.

nnually, more than 10,000 people in the United States

sustain a spinal cord injury1. About 15% of these injuries are burst fractures2, which occur predominantly in
younger patients, incurring a high financial and societal cost3.
During the fracture process, one or more fragments of
the vertebral body are retropulsed into the spinal canal, and
many surgeons advocate decompressive surgery, even in patients with otherwise clinically stable fractures, to prevent or
reduce a neurological deficit4. Computed tomography demonstrates the degree of spinal stenosis, but there is doubt as to
whether the position of the fragments seen on such scans represents the true extent of the canal occlusion produced during
the fracture process5. There is also evidence that, after a fracture, any ongoing compression of the spinal cord has no effect
in patients with a stable neurological status and that some recovery of function can be expected even if the patient is
treated nonoperatively4,6.

The hypothesis for this study was that the final resting
positions of the bone fragments seen on computed tomography imaging are not representative of the dynamic fracture
process. Implicit in this hypothesis is the assumption that the
damage to the spinal cord occurs at the moment of injury. In
previous experimental studies, investigators have used indirect
measurement techniques to determine the occlusion of the
spinal canal after the removal of the spinal cord5,7. By using a
combined experimental and computational approach, we
were able to investigate the dynamic fracture process by direct
measurement of the extent of canal occlusion and by simulation of spinal cord impact.
Materials and Methods
Experimental Model
horacolumbar spinal specimens from twenty-one-dayold male Holstein calves were retrieved from an abattoir


THE JOUR NAL OF BONE & JOINT SURGER Y JBJS.ORG
VO L U M E 85-A N U M B E R 11 N O VE M B E R 2003

Fig. 1-A

A DYNAMIC STUDY

OF

T H O R A CO L U M B A R B U R S T F R A C T U R E S

Fig. 1-B

Experimental apparatus for drop-weight tests showing the setup for the high-speed video experiments (Fig. 1-A) and the pressure measurements (Fig. 1-B).

and frozen at 20C. After defrosting for twenty-four hours,

the specimens were cut into three-vertebra segments, excess
paravertebral muscle was removed, and the spinal cord was
extracted from the spinal canal. The ends of the specimen
were set in 80-mm-diameter polymethylmethacrylate end
plates to produce flat, parallel surfaces. During casting, the
ends of the spinal canal were kept free of cement so that it was
possible to view the inside of the canal during testing.
Burst fractures were produced with use of the dropweight method8, in which a 1 to 7-kg mass was dropped axially

onto the specimen from a height of 2 m. This allowed the

impact energy to be varied while the impact velocity was
maintained. The specimens were housed between two stainless-steel end plates, with the top plate guided to permit movement only in the impact direction.
High-Speed Video Measurements
The first series of tests was carried out with use of high-speed
video imaging down the spinal canal during impact. A total of
twenty-one specimens were tested at impact energies ranging

Fig. 2

Finite element model at the initial point of

fragment projection. Axial compression of the
model had previously been simulated to the
point of osseous fracture. PLL = posterior
longitudinal ligament.


THE JOUR NAL OF BONE & JOINT SURGER Y JBJS.ORG
VO L U M E 85-A N U M B E R 11 N O VE M B E R 2003

A DYNAMIC STUDY

OF

T H O R A CO L U M B A R B U R S T F R A C T U R E S

Fig. 3

Typical graph of canal occlusion versus time

as well as representative images at 0, 4, 10,
and 20 msec after impact.

from 60 to 140 J. The method and its validation have been described in detail previously9. Briefly, each steel end plate contained a cutaway section in which a mirror was positioned at
45 to the impact direction (Fig. 1-A). The specimen was positioned so that the spinal canal was aligned between the two
mirrors. Light was shone by means of the top mirror through
the spinal canal and reflected by the bottom mirror into the
lens (f/2.8D, AF 60 mm; Nikon, Melville, New York) of the
high-speed camera (Kodak HS4540; Roper Scientific, MASD,
San Diego, California) running at 4500 frames per second.
This allowed silhouette images of the cross section of the spinal canal to be captured on video during the impact. After
each test, the images were downloaded to a personal computer
and were subsequently analyzed with use of proprietary software (Image Pro Plus 3.0; Media Cybernetics, Silver Spring,
Maryland). A custom-written algorithm was used to trace the
outline of the open canal, with the change in light intensity as
the canal was occluded taken into account. The cross-sectional
area of the canal was then determined for each frame. Initial
validation tests employing a second high-speed video camera

Fig. 4-A

to film the outside of the specimen were performed to determine if there was any lateral movement or buckling of the
specimen during impact. After impact, the nine specimens
used in the validation tests were also imaged with computed
tomography. The degree of canal occlusion was calculated as
the ratio of the postimpact canal area to that of the average canal area of the two adjacent vertebrae.
Pressure Measurements
A second series of experiments was carried out to determine
the increase in pressure in a synthetic spinal cord during the
fracture process. A total of twenty-one tests were performed at
impact energies ranging from 20 to 140 J. A catheter-tip pressure transducer was custom manufactured (Gaeltec, Dunvegan, Isle of Skye, United Kingdom) and was connected to a
data acquisition system (LabVIEW 5.0; National Instruments,
Austin, Texas). The frequency response of the system was determined, and the 3-dB cutoff was found to be in excess of 8
kHz with a negligible ripple band below that frequency. The
impact force was measured with a quartz load cell (Type 904B;

Fig. 4-B

Graphs of energy of impact versus maximum (Fig. 4-A) and final (Fig. 4-B) levels of canal occlusion as measured from the high-speed video.
Error bars show the standard deviation.


THE JOUR NAL OF BONE & JOINT SURGER Y JBJS.ORG
VO L U M E 85-A N U M B E R 11 N O VE M B E R 2003

Fig. 5-A

A DYNAMIC STUDY

OF

T H O R A CO L U M B A R B U R S T F R A C T U R E S

Fig. 5-B

Mean-difference graphs of computed tomography-measured occlusion and maximum video-measured occlusion (Fig. 5-A) and computed
tomography-measured occlusion and final video-measured occlusion (Fig. 5-B). The mean error and the 95% confidence interval are shown as
dotted and dashed lines, respectively.

Kistler Instrumente, Winterhur, Switzerland) housed in the

lower end plate. The output was again fed into a personal
computer by means of a charge amplifier. The 3-dB cutoff for
the load cell system was in excess of 50 kHz.
To measure the pressure, the inferior end of the spinal
canal was sealed with modeling clay and the transducer was
suspended vertically in the canal from the superior end of the
specimen. The sensing pad was aligned anteriorly at the central vertebra, and the transducer was held in place with a
metal clip. There is a rapid deterioration in the mechanical
properties of bovine spinal cord with freezing and time post
mortem10 so a synthetic cord material, prepared from 9%
(weight/weight) gelatin solution, was used for all tests. A previous study showed that, at this concentration, the response of
gelatin under lateral impact is similar to that of a fresh spinal
cord11. The solution was prepared with use of distilled water
and type-A gelatin from porcine skin (G2500; Sigma-Aldrich,
Poole, Dorset, United Kingdom) and was poured into the canal. Once the solution had gelled, the metal clip was removed
so that the transducer stem was in contact only with the gelatin (Fig. 1-B). This prevented vibrations from the surrounding
hard tissue from being transmitted to the sensing pad through
the transducer stem.
After testing, all of the specimens were imaged with
computed tomography and the degree of canal occlusion was
determined as before.
Finite Element Analysis
The effect of the spinal cord on the dynamics of the fragment
were further investigated with use of computational simulation. A finite element model of a three-vertebra segment had
previously been constructed from the computed tomography
scans of a typical bovine specimen, and this was used to simulate the impact on the specimen up to the moment of vertebral body impact12. The geometry of the specimen at the time
of fracture was used to simulate fragment retropulsion by ap-

plying a velocity to the bone fragment in the posterior direction (Fig. 2). The model included the spinal cord, dura mater,
posterior elements, and posterior longitudinal ligament, which
had become slack as a result of the compression of the vertebral body. The material properties were derived from the literature (Table I). Different fragment velocities ranging from 1 to
20 m/sec were simulated to encompass the range that occurred in the experimental tests. Each simulation was repeated
with the spinal cord and dura mater removed, and the difference in fragment trajectories was determined.
Results
High-Speed Video Measurements
n every case, the images obtained from the high-speed
video clearly showed the vertebral bone fragment being
projected into the spinal canal and then recoiling to the final
resting position. The second camera showed no visible buckling of the specimen during impact and minimal lateral movement, giving a mean estimated error (and standard deviation)
of 7% 4% for the calculated cross-sectional area of the spinal
canal9. The relative cross-sectional area of the canal was plotted as a function of time (Fig. 3), and in every case the final
level of canal occlusion was less than the maximum that occurred during impact. There was also an increase in maximum occlusion with an increase in impact energy (Fig. 4-A).
Repeated tests at the same impact conditions showed similar
levels of maximum occlusion, but there was greater variability
in the final fragment position (Fig. 4-B).
The images of the nine specimens that were scanned
with computed tomography showed that, in every case, a
burst fracture had been produced, with lower impact energies
(up to 60 J) producing predominantly Denis2 type-C fractures
and higher energies producing more type-A fractures. The
fracture patterns were comparable with those seen in the human spine. Dissection of the central vertebra of each specimen
also revealed a trapezoidal or wedge-shaped fragment typical


THE JOUR NAL OF BONE & JOINT SURGER Y JBJS.ORG
VO L U M E 85-A N U M B E R 11 N O VE M B E R 2003

A DYNAMIC STUDY

OF

T H O R A CO L U M B A R B U R S T F R A C T U R E S

TABLE I Mechanical Properties of Soft Tissues Used in Finite Element Study

Material

Properties

References

Posterior longitudinal ligament

E = 18.5 MPa ( < 11%)

E = 61.6 MPa (11% < < 34%)
E = 46 MPa ( > 34%)

Dura mater

Anisotropic
E = 142 MPa (circumferential/radial)
E = 0.7 MPa (longitudinal)

18

Spinal cord

E = 1.3 MPa

19

of a burst fracture, with no obvious interference from the ring

apophysis. The level of agreement between the canal occlusion
measured on the postfracture computed tomography scans
and the maximum dynamic occlusion seen on the high-speed
video images was assessed with use of statistical methods proposed by Lin13. The concordance coefficient was 0.483 (95%
confidence interval, 0.164 to 0.802), which was significantly
different from the concordance coefficient between the occlusion measured with computed tomography and the final occlusion of 0.938 measured from the video (95% confidence
interval, 0.793 to 0.982) (Figs. 5-A and 5-B).
Canal Pressure
In all of the tests, the pressure was seen to rise to a maximum
and return to zero. The mean total length of the pulse was 42
15 msec, with no dependence on the impact energy (Pearson
correlation = 0.01). There was a rise in the mean maximum
pressure with increasing impact energy (r = 0.659, p = 0.001).
There was a high correlation between the maximum pressure
and the peak load measured on the bottom load cell (r =
0.937, p < 0.001). The computed tomography images showed
that all impacts of >20 J had produced burst fractures, but
there was not a significant correlation, with the numbers
available, between the extent of occlusion seen on the image
and the impact energy, peak pressure, or peak load (Pearson
correlation coefficient = 0.414 0.062, 0.349 0.121, and
0.364 0.105, respectively).
Finite Element Analysis
The finite element analysis showed that, at high fragment velocities, the presence of the spinal cord and dura mater made a
considerable difference to the fragment propagation, with

16, 17

these tissues acting both to reduce the maximum displacement and to increase the level of recoil (Table II).
Discussion
he calf spine has been shown to exhibit mechanical responses similar to those of the human spine under a range
of loading conditions14, and it has been used previously to
model the burst fracture process15. In the current study, bovine
specimens were chosen to reduce interspecimen variability
since they could be harvested from animals with a narrow age
range. The fracture patterns that were produced corresponded
well with those observed in clinical practice, and the model was
therefore considered to be representative of the human spine.
The impact energies were within the range of those used in cadaveric studies5,7,8, although a comparison of equivalent fractures suggests that slightly higher energies were required for the
bovine specimens because of the higher density of bovine bone.
The low level of concordance between the amount of canal occlusion measured on the postfracture computed tomography
scans and the maximum occlusion demonstrated by the video
at impact indicates that the computed tomography image did
not represent the maximum dynamic canal occlusion and
therefore cannot be used to determine the events that occurred
during the fracture. At higher-impact energies, the entire posterior section of the vertebral body became detached during the
fracture process. In some cases, the lack of constraint on the
fragment allowed it to rotate and become lodged in the canal. In
other cases, alignment with the vertebral body was maintained
and the fragment was recoiled back into the vertebral body. It
was therefore not possible to predict solely from the final position of the fragment what degree of canal occlusion had occurred dynamically.

TABLE II Difference in Simulated Maximum Lateral Fragment Displacement with and without the Spinal Cord
Maximum Fragment Displacement (mm)
Speed (m/sec)

With Cord and Dura

Without Cord and Dura

Difference Between
Displacements
(% of canal diameter)

3.7

4.2

3.6

4.3

6.1

11.9

4.8

6.9

14.2

10

5.4

8.0

17.6


THE JOUR NAL OF BONE & JOINT SURGER Y JBJS.ORG
VO L U M E 85-A N U M B E R 11 N O VE M B E R 2003

The high level of correlation between the pressure readings and impact conditions (r = 0.937 for peak impact load)
demonstrates that the pressure produced in the spinal canal
depends on the magnitude of the impact. The duration of the
pressure pulse was similar to that seen on the load cell below
the specimen and was considerably longer than that of the
fragment motion across the canal observed on the high-speed
video. This indicates that the pressure rise is the summation of
the stress wave generated by the impact on the specimen and
that produced locally by the fragment impact. However it was
caused, the maximum pressure generated in the spinal canal is
likely to be a good indicator of the level of spinal cord damage
and hence the neurological deficit. The low correlation between maximum pressure and the degree of occlusion measured on the computed tomography scan (r = 0.349) indicates
that a computed tomography measurement of occlusion after
a fracture cannot be used to determine the dynamic rise in
spinal canal pressure during the impact and thus cannot be
used to indirectly measure the extent of spinal cord damage.
Since the spinal cord and dura mater could not be left in
place during the high-speed video experiments, a finite element
model was used to simulate their effect on the fragment dynamics. The model showed that at high fragment velocities, as would
occur at 140-J impacts, the presence of the cord and dura made a
considerable difference with regard to the amount of canal occlusion that occurred. The cord and dura appeared to act in a
manner similar to that of the posterior longitudinal ligament in
reflecting the fragment back toward the vertebral body. Hence,
the overall pattern of the fragment dynamics is the same with or
without the cord, with the fragment being projected into the spinal canal before recoiling back toward the vertebral body.
Some surgeons operate on patients when computed tomography images demonstrate canal compromise of more than
a fixed amountusually 40% or 50%. This amount is chosen

A DYNAMIC STUDY

OF

T H O R A CO L U M B A R B U R S T F R A C T U R E S

by surgeons on the basis of anecdotal evidence rather than controlled clinical studies4. Both the experimental and the computational results of our study showed that the final level of
occlusion does not represent the greatest occlusion that occurs
during impact. Furthermore, the results of the high-speed video
tests showed that, at higher levels of occlusion, the final position
of the fragment was poorly correlated with the maximum level
of impingement. Any neurological damage is likely to occur at
the point of maximum canal occlusion, which also corresponds
with the maximum pressure generated in the cord. Hence, spinal cord damage probably is not accurately represented by the
final resting positions of the bone fragments seen on postinjury
imaging of the spinal canal.

Ruth K. Wilcox, PhD
David C. Barton, PhD
Richard M. Hall, PhD
School of Mechanical Engineering, University of Leeds, Leeds LS2 9JT,
United Kingdom. E-mail address for R.K. Wilcox: r.k.wilcox@leeds.ac.uk
Thomas O. Boerger, FRCS
David J. Allen, FRCS
David Limb, BSc, FRCSEd(Orth)
Robert A. Dickson, DSc
Musculo-Skeletal Services, CSB, St. Jamess Hospital, Leeds LS9 7TF,
United Kingdom
In support of their research or preparation of this manuscript, one or
more of the authors received grants or outside funding from the Wishbone Trust, Yorkshire Childrens Spine Foundation, and the Engineering
and Physical Sciences Research Council. None of the authors received
payments or other benefits or a commitment or agreement to provide
such benefits from a commercial entity. No commercial entity paid or
directed, or agreed to pay or direct, any benefits to any research fund,
foundation, educational institution, or other charitable or nonprofit
organization with which the authors are affiliated or associated.

References
1. Go BK, DeViro MJ, Richards JS. The epidemiology of spinal cord injury. In:
Stover SL, DeLisa JA, Whiteneck GG, editors. Spinal cord injury: clinical outcomes from the model systems. Gaithersburg, MD: Aspen; 1995. p 21-5.
2. Denis F. The three column spine and its significance in the classification of
acute thoracolumbar spinal injuries. Spine. 1983;8:817-31.
3. DeVivo MJ. Causes and costs of spinal cord injury in the United States. Spinal Cord. 1997;35:809-13.
4. Boerger TO, Limb D, Dickson RA. Does canal clearance affect neurological
outcome after thoracolumbar burst fractures? J Bone Joint Surg Br. 2000;82:
629-35.
5. Chang DG, Tencer AF, Ching RP, Treece B, Senft D, Anderson PA. Geometric
changes in the cervical spinal canal during impact. Spine. 1994;19:973-80.
6. Limb D, Shaw DL, Dickson RA. Neurological injury in thoracolumbar burst
fractures. J Bone Joint Surg Br. 1995;77:774-7.
7. Panjabi MM, Kifune M, Wen L, Arand M, Oxland TR, Lin RM, Yoon WS,
Vasavada A. Dynamic canal encroachment during thoracolumbar burst fractures. J Spinal Disord. 1995;8:39-48.
8. Willen J, Lindahl S, Irstam L, Aldman B, Nordwall A. The thoracolumbar
crush fracture. An experimental study on instant axial dynamic loading: the
resulting fracture type and its stability. Spine. 1984;9:624-31.
9. Wilcox RK, Boerger TO, Hall RM, Barton DC, Limb D, Dickson RA. Measurement of canal occlusion during the thoracolumbar burst fracture process.
J Biomech. 2002;35:381-4.
10. Oakland RJ, Wilcox RK, Hall RM, Barton DC. The mechanical response of
spinal cord to uniaxial loading. Trans Orthop Res Soc. 2002;27:783.

11. Wilcox RK, Boerger TO, Hall RM, Barton DC. Canal pressure measurements
and video recording of thoracolumbar burst fractures. Presented as a poster
exhibit at the Annual Meeting of the American Academy of Orthopaedic Surgeons; 2001 Feb 28-Mar 4; San Francisco, CA.
12. Wilcox R, Allen D, Barton D, Hall R, Limb D, Dickson R. An investigation
of the burst fracture mechanism using a combined experimental and finite
element approach. Read at the Fourth World Congress of Biomechanics
and American Society of Biomechanics; 2002 Aug 4-9; Calgary, Alberta,
Canada.
13. Lin LI. A concordance correlation coefficient to evaluate reproducibility. Biometrics. 1989;45:255-68.
14. Wilke HJ, Krischak S, Claes L. Biomechanical comparison of calf and human
spines. J Orthop Res. 1996;14:500-3.
15. Tran NT, Watson NA, Tencer AF, Ching RP, Anderson PA. Mechanism of the
burst fracture in the thoracolumbar spine. The effect of loading rate. Spine.
1995;20:1984-8.
16. Tkaczuk H. Tensile properties of human lumbar longitudinal ligaments. Acta
Orthop Scand. 1968;Suppl 115:1+.
17. Chazal J, Tanguy A, Bourges M, Gaurel G, Escande G, Guillot M, Vanneuville
G. Biomechanical properties of spinal ligaments and a histological study of
the supraspinal ligament in traction. J Biomech. 1985;18:167-76.
18. Wilcox R, Barton D, Bilston L, Hall R. A mathematical model for the viscoelastic properties of dura mater. Trans Orthop Res Soc. 2001;26:708.
19. Bilston LE, Thibault LE. The mechanical properties of the human cervical
spinal cord in vitro. Ann Biomed Eng. 1996;24:67-74.