SPINE Volume 28, Number 19, pp 2204–2208
©2003, Lippincott Williams & Wilkins, Inc.
Failure of Human Cervical Endplates: A Cadaveric
Experimental Model
Eeric Truumees, MD,* Constantine K. Demetropoulos, PhD,† King H. Yang, PhD,‡ and
Harry N. Herkowitz, MD‡
Study Design. An in vitro biomechanical study using a
servohydraulic testing machine on cadaveric endplates.
Objectives. To characterize the effects of bone mineral
density, endplate geometry, and preparation technique
on endplate failure load.
Summary of Background Data. The effects of endplate
preparation methods on failure loads are only partly char-
acterized in the literature. Endplate burring has been recom-
mended to increase fusion rates. However, graft subsidence
may complicate anterior reconstruction procedures.
Methods. After radiographic screening, 21 cadaveric
cervical spines underwent dual-energy x-ray absorptiom-
etry scanning to quantify mineral content. Endplate ge-
ometry was calculated in 55 randomly selected endplates
from the inferior C2 to the superior T1 levels. These ver-
tebrae were embedded in polyester resin and randomly
left intact, perforated, or burred. The cervical endplates
were loaded at a rate of 0.2 mm/s on an Instron materials
tester with an attached 9 mm diameter polycarbonate rod
(an area of 64 mm2). A stepwise, univariate linear regres-
sion was used to compare the point of endplate failure
with the vertebral level, endplate area, gender, age, bone
mineral density, and preparation technique.
Results. Mean bone mineral density, as measured by
dual-energy x-ray absorptiometry, was 0.713 g/cm2 (⫾
0.173 g/cm2). Mean endplate area was calculated at 323
mm2. A mean compressive force of 754 N (⫾ 445 N) was
required before endplate failure. Trends toward increas-
ing compressive loads were noted with decreasing end-
plate area and increasing bone mineral density. Increas-
ing age (P ⫽ 0.0203), caudal vertebral level (P ⬍ 0.0001),
endplate burring (P ⫽ 0.0068), and female gender (P ⫽
0.0452) were associated with significantly lower endplate
fracture loads in compression.
Conclusions. Bone quality was predictive of endplate
compressive failure loads. Intact endplates failed at sig-
nificantly higher loads than their perforated or burred
counterparts. [Key words: cervical spine, biomechanics,
From the †Department of Orthopaedic Surgery, *William Beaumont
Hospital, Royal Oak, and ‡Bioengineering Center, Wayne State Uni-
versity, Detroit, Michigan.
This study was funded, in part, by an unrestricted grant from
Medtronic-Sofamor-Danek.
This paper was presented, in part, at the 28th Annual Meeting of the
Cervical Spine Research Society; 2000; Charleston, South Carolina.
Acknowledgment date: April 15, 2002. First revision date: June 28,
2002. Second revision date: September 18, 2002. Acceptance date:
January 24, 2003.
The manuscript submitted does not contain information about medical
device(s)/drug(s).
Corporate/Industry funds were received to support this work. No ben-
efits in any form have been or will be received from a commercial party
related directly or indirectly to the subject of this manuscript.
Address correspondence to Christine Musich, 3535 West 13 Mile
Road, Suite 605, Royal Oak, MI 48073, USA; E-mail:
Truumees@Comcast.net
2204
graft, anterior interbody fusion, endplate, bone density,
compressive strength, postoperative complications]
Spine 2003;28:2204 –2208
Anterior cervical discectomy and corpectomy techniques
are commonly employed in the treatment of radiculopa-
thy and myelopathy. Occasionally, these procedures are
complicated by graft collapse or extrusion, endplate fail-
ure and subsidence, or nonunion.1 Presently, the optimal
compressive force to decrease bone graft extrusion and
encourage fusion is not known.2,3,4,5
The ability of various grafts to withstand compressive
loading has been characterized. White and Hirsh tested
various grafts employed in anterior cervical spine sur-
gery. 6 They found that Smith-Robinson tricortical
wedges were significantly stronger (344 kp) than other
graft configurations (188 –195 kp). The authors esti-
mated that these grafts would be subjected to 6.3 kg of
force. They did not include postoperative compressive
forces from muscle tension or neck flexion.
Wittenberg et al studied compressive strengths of var-
ious grafts under axial compression.7 In this study, fibu-
lar strut grafts withstood 5070 N (⫾ 3.250 N) of axial
loading before failure. Failure was defined as the maxi-
mum load on the load-displacement curve. Fibular strut
grafts were significantly stronger than anterior (1150 ⫾
487 N) and posterior (667 ⫾ 311 N) iliac crest and rib
(452 ⫾ 192 N) grafts.
An et al used six fresh frozen cadaveric pelves to assess
the potential usefulness of dual-energy x-ray absorpti-
ometry (DEXA) to predict the strength of iliac crest bone
grafts.8 They reported that the biomechanical strength of
iliac bone is linearly related to its bone mineral density
(BMD) as measured by DEXA. Their Smith-Robinson
grafts failed at 1368 N. Anterior iliac crest strut grafts
failed at 2168 N.
Endplate failure loads have not been as well charac-
terized. In one study of an entire motion segment, the
cervical vertebral body could withstand compressive
loads of 1310 N.4 Lim et al studied endplate preparation
techniques and found that intact endplates failed at a
mean compressive load of 634 N.9 Partly burred end-
plates failed at 494 N, whereas completely burred end-
plates failed at 419 N. These authors used a finite ele-
ment model to assess the effects of endplate perforation
on endplate failure loads. The effect of one large central
hole was compared to two- and four-hole configura-
tions. They found that one large central hole was prefer-
 Cervical Endplate Compression Load • Truumees et al 2205

Table 1. Distribution of Cervical Spine Specimens

Sample Area Failure BMD Age
Preparation Size (mm2) Load (N) (g/cm2) (yrs)

Intact 35 314 ⫾ 106 897 ⫾ 453 0.723 ⫾ 198 72 ⫾ 13

Perforated 10 348 ⫾ 111 590 ⫾ 347 0.681 ⫾ 108 74 ⫾ 16
Burred 10 326 ⫾ 101 403 ⫾ 208 0.676 ⫾ 119 70 ⫾ 16
BMD ⫽ bone mineral density.

able in that it better distributed compressive load across

the remaining intact endplate.
We postulate that three variables affect load to failure
of cervical endplates in anterior fusion procedures: end-
plate geometry, bone quality, and endplate preparation
technique. Endplate geometry was characterized based
on vertebral level and endplate area. Bone quality was
characterized by patient age, gender, and BMD. End-
plates were left intact or were prepared for fusion by
perforation or burring.
Materials and Methods
A total of 28 fresh frozen human cadaveric cervical spines were
obtained through the willed body program at the University of
Michigan and the International Institute for the Advancement
of Medicine. Sixty percent were female. The average age at Figure 1. Diagram of the perforation pattern employed.
demise was 72 years (⫾ 14 years). Anteroposterior (AP) and
lateral radiographs were obtained to exclude lytic lesions or
other anomalies. A DEXA scan was then obtained to quantify test subjects were then loaded at a rate of 0.2 mm/s on an
mineral content. Average as well as level specific, lateral BMD Instron materials tester (Canton, MA (Figure 2). Endplate fail-
values were obtained using a Lunar DPXL system (Madison, ure was defined as reversal of the force-displacement curve
WI). After screening, 21 cadavers were selected for further wherein decreasing force was needed to further advance the
study. The remaining seven specimens were found to have prior rod. For the burred specimen, clear identification of the failure
cervical surgery, marked degeneration to the point of autofu- point was difficult. Each specimen was compressed to 5 mm.
sion, lytic lesions, or other anomalies. Failure was defined as that load at which the indenter advanced
Fifty-five vertebrae were harvested from the 21 spines. Lev- greater than 0.25 mm without further increase in load. A step-
els were randomly selected and included the inferior endplate wise linear regression was then used to compare failure load
of C2 to the superior endplate of T1. Please see Table 1 for with bone mineral density, gender, age, vertebral level, end-
individual specimen characteristics. Individual vertebrae were plate geometry, and endplate preparation technique.
separated, and all soft tissue was removed. Bony elements were
Results
left intact, including the posterior elements. The specimens
were then potted in a polyester resin to the approximate mid- Mean BMD in these specimens was 0.713 g/cm2 (⫾
point of the vertebral body in order to maintain a perpendicular 0.173 g/cm2). The male specimens had a greater mean
orientation of the upper endplate relative to the indenter. End-
plate geometry was calculated using a direct caliper measure-
ment of the outer vertebral width and length. Area was calcu-
lated assuming elliptical geometry.
The cartilage endplate was removed using a curette. Thirty-
five specimens were tested without additional endplate prepa-
ration (the intact specimen). Given that the indenter was stan-
dardized in size, the perforations were standardized relative to
the indenter, not the size of the vertebral endplate. Their con-
figuration was meant to simulate surface area for vascular in-
flow into the center (noncortical) portion of the strut graft. The
perforations were made with a 1-mm sharp, fine curette in a
square pattern centered on the middle of the docking site for
the indenter. The perforations were 3 mm apart (center to cen-
ter) (Figure 1). In the burred specimens, the entire visible end-
plate medial to the uncovertebral joints was gradually burred Figure 2. A polycarbonate rod was fashioned to simulate the
until the underlying trabecular bone became visible. geometry of an intervertebral strut graft. This was then used to
A 9 mm diameter (area 64 mm2) polycarbonate rod was compress into the vertebral endplate using an Instron materials
fashioned to simulate a typical intervertebral strut graft. The tester.
2206 Spine • Volume 28 • Number 19 • 2003

BMD (0.8133 ⫾ 0.201 g/cm2) than did the female (0.648

⫾ 0.114 g/cm2) (P ⫽ 0.0193). However, the male spec-
imens were also, on average, younger (66.2 vs. 75.8
years). Finally, males had a greater mean maximum com-
pressive failure load than females (911 vs. 650 N). Al-
though these differences are significant, we cannot com-
ment on which variable, age or gender, had the greater
impact in that the male specimens were also younger.
A mean 754 N (⫾ 445 N) force was required before
endplate failure occurred. In the intact specimen, failure
occurred abruptly (at a mean 897 N) as the cortical end-
plate gave way (Figure 3A). There was continued resis-
tance to rod advancement, however, through compres-
sion of the underlying cancellous bone. Perforated
specimens demonstrated a similar pattern of vertebral
penetration at a lower mean force, 590 N (Figure 3B).
Burred specimens exhibited a gradual failure with slow
impaction of the subendplate bone (Figure 3C). Failure
occurred with less force than in either the perforated or
intact specimens, at 403 N. Specimen age and BMD were
similar between the endplate preparation groups: intact
(72 years, 0.732 g/cm2), perforated (74 years, 0.681
g/cm2), and burred (70 years, 0.676 g/cm2).
A stepwise, univariate linear regression was used to
compare the point of endplate failure with the vertebral
level, endplate area, gender, age, BMD, and preparation
technique. A trend of increasing endplate failure load
was noted with increasing endplate area. Similarly, in-
creasing BMD was associated with increasing endplate
failure load, but significance was not reached. Increasing
age (P ⫽ 0.0203), vertebral level below C5 (P ⬍ 0.0001),
endplate burring (P ⫽ 0.0068), and female gender (P ⫽
0.0452) were associated with statistically significant
lower endplate compressive loads (Figures 4 and 5).

Discussion Figure 3. A, A typical load-displacement curve from a specimen in

We theorized that ultimate endplate compressive load
could be predicted based on three variables: endplate
geometry, bone quality, and endplate preparation tech-
nique. Endplate geometry may affect stress distribution.
Axial force transmitted closer to the cortical rim of the
vertebral body may allow increased loads before end-
plate failure. However, in our study, endplate dimen-
sions did not significantly predict the ability of the end-
plate to resist failure when compressed with a
standardized, simulated strut graft.
Inferior bone quality, as measured by increasing age
and female gender, was associated with lower compres-
sive loads at failure. Younger and male specimens had
stronger endplates as measured by higher endplate fail-
ure loads. Female gender may also affect vertebral geom-
etry; however, this was not directly measured. Bone min-
eral density measures, via DEXA scanning, did not
predict endplate compressive loads in a statistically sig-
nificant manner. We feel that more accurate measures of
BMD in these spondylotic, elderly cadavers may have
produced significant results. Cervical DEXA lacks a
which the endplate was left intact. That is, after cartilage removal,
no further perforation or burring was undertaken. In this specimen,
sudden failure of the endplate occurred at 650 N. This was fol-
lowed by continued resistance to further compression by the
subcortical trabecular bone. Failure is indicated by an “X.” End-
plate failure loads were significantly different in each group. That
is, intact endplates failed at a significantly higher load than did
perforated (P ⬍ 0.023). Similarly, perforated endplates failed at a
significantly higher load than did burred (P ⬍ 0.017). When intact
endplates are compared with burred endplates, the difference is
great (P ⫽ 0.0068). B, A typical load-displacement curve of a
cervical endplate after perforation with a curette. Although this
curve is similar to that of the intact endplate, the failure occurs at
a lower load, 418 N. Moreover, the loading curve does not dem-
onstrate the sharp demarcation noted with failure of the intact
endplate. Failure is indicated by an “X.” C, A typical load-
displacement curve of a cervical vertebra after endplate removal
with a burr. In this case, a gradual increase in load occurs with
increasing displacement as the trabecular bone is compacted. A
discrete point of failure is difficult to identify, but was selected
based on a 0.25 mm displacement without a significant increase in
load; in this example, 234 N. Failure is indicated by an “X.”

Figure 4. The effects of increasing endplate violation on load at
failure. These differences were statistically significant.
readily available, specific scanner protocol and may be
inaccurate in spondylotic specimens.10
The method of preparation did significantly affect the
endplates ability to withstand compressive loads. End-
plate preparation, such as perforation and burring, are
often used to encourage graft incorporation.11,12 How-
ever, intact endplates failed at significantly higher loads
than their perforated or burred counterparts.
Despite differences in experimental design, our values
for endplate failure loads of intact and burred specimens
are similar to previously reported values.9 Our study dif-
fers from previous work in that the vertebrae were not
sectioned in the horizontal plane, but were tested intact.
Ours is the first study to directly measure the failure load
of perforated endplates. In our study, endplates were
compressed to a depth of 5 mm; this additional depth
was selected to ensure complete failure of our specimens.
Particularly in the burred specimens, variability in the
pattern of trabecular failure may lead to partial collapse
being falsely recorded as complete failure.
Although Lim et al found that endplate thickness was
not directly related to failure load, greater degrees of
endplate burring did correspond to lower failure loads.9
These findings imply that endplate integrity, rather than
Figure 5. Male endplates fail at a higher mean load than female.
This difference was statistically significant.
Cervical Endplate Compression Load • Truumees et al 2207
absolute thickness, is critical in load bearing. In Lim et
al’s study, a finite element model was used to assess the
effect of endplate perforation. These authors found that a
single, larger central perforation decreased endplate
strength less than multiple, smaller perforations. In our
study, we directly tested the effect of four small endplate
perforations. Although these perforations led to a signif-
icantly lower endplate failure load, this failure load was
significantly higher than that achieved after burring the
endplate. In clinical practice, we prefer not to create one
larger hole centrally because exact geometry and place-
ment of the intervertebral graft may vary. End-bearing of
the structural portion of the graft on the hole may lead to
subsidence. The size and distribution of smaller perfora-
tions may be more easily altered from case to case as
bone quality and geometry dictate.
We compared endplate failure loads from our study
and those of Lim et al with graft failure data from the
reports by An et al, White and Hirsh, White et al, and
Wittenberg et al.6,7,8,13 When loaded in compression,
strut grafts were between 1.3 and 8 times stronger than
the intact endplate. Smith-Robinson grafts were 1.3 to
5.3 times stronger than the intact endplate. Strut grafts
were much stronger than burred endplates (5.4 to 12.6
times). Smith Robinson grafts were from 2.8 to 8.4 times
stronger than a burred endplate.
In a separate study, we measured the compressive load
on a simulated Smith Robinson anterior cervical graft.14
The mean compressive load on a 6 mm graft loaded with
simulated head weight was 16.2 N. The intact endplate
withstood 55.4 times these measured loads. In the same
study, a larger, 8 mm graft was subjected to a compres-
sive load of 29.2 N, or one-thirtieth of the intact endplate
failure loads measured in this study. Burred endplates
were 13.8 to 24.9 times stronger than these ex vivo mea-
surements of compressive load on 6 or 8 mm grafts. This
study of anterior strut grafts simulated loading at the
time of graft insertion before postoperative graft com-
pression due to muscle forces and neck flexion.
Although in vivo cervical graft loads have not been
extensively described in the literature, Moroney et al and
others have estimated compressive loads on entire cervi-
cal motion segments using a mathematical model and
electromyograph (EMG) data.15 In flexion, the compres-
sive load at C4 –C5 was estimated at 558 N. This com-
pressive load was therefore less than the failure load of
an intact endplate (754 N). However, burred endplates
failed at lower compressive loads (403 N) than those
estimated based on EMG data. These endplates would
therefore be subject to failure and graft subsidence. In
our study, perforated endplates fail at only slightly
higher loads (590 N vs. 558 N) than those estimated by
Moroney et al.
Our study is limited by its ex vivo assessment of bone
strength. The differences in mechanical properties of
bone in vivo are not known. Our use of a polycarbonate
indenter only approximates axial force delivered by a
strut graft in an anterior cervical fusion. Moreover, dur-
2208 Spine • Volume 28 • Number 19 • 2003
ing the initial period of graft incorporation, vascular in-
growth is associated with a decrease in mechanical
strength. Later, this strength returns.16 Our model esti-
mates failure load at the time of surgery only. Ultimately,
the factors selected were able to account for approxi-
mately 48% of the observed variability in endplate fail-
ure load. These elderly, cadaveric specimens represent a
worst-case scenario.
Endplate preparation technique significantly affects
failure load. Although true in vivo loads are not well
characterized, they likely exceed values for burred end-
plates in osteoporotic patients. Whereas failure loads of
perforated endplates are at or near in vivo compressive
loads, intact endplates typically bear greater compressive
force before failure. Therefore, in patients with signifi-
cant osteoporosis, less aggressive methods of endplate
preparation and longer term bracing should be consid-
ered. Perhaps sizing the graft such that load-bearing
edges lay near the anterior and posterior margins of the
vertebral body will increase endplate-bearing capacity.
Key Points
● Endplate characteristics predicting decreasing
failure load include increasing patient age and fe-
male gender.
● More aggressive endplate preparation tech-
niques such as endplate burring significantly de-
crease failure loads.
● The cervical endplate fails at loads less than
those typically measured for compressive failure of
the cervical graft.
Acknowledgments
The authors thank Mamtha Balasubramaniam, Biostat-
istician, William Beaumont Hospital, and Stuart Dees,
Supervisor, Department of Nuclear Medicine, William
Beaumont Hospital.
References
1. Robinson R, Walker AE, Ferlic D, et al. Results of anterior interbody fusion
of the cervical spine. J Bone Joint Surg Am 1962;44:1569 – 87.
2. An HS, Evanich C, Nowichi B, et al. Ideal thickness of Smith-Robinson
anterior cervical interbody graft. Spine 1993;18:2043–7.
3. Kazarian L, Graves GA. Compressive strength characteristics of the human
vertebral centrum. Spine 1977;2:1–14.
4. Moroney SP, Schultz AB, Miller JAA, et al. Load-displacement properties of
lower cervical spine motion segments. J Biomech 1988;21:769 –79.
5. Olsewski JM, Garvey TA, Schendel MJ. Biomechanical analysis of facet and
graft loading in a Smith-Robinson type cervical spine model. Spine 1994;19:
2540 – 4.
6. White AA, Hirsh C. An Experimental study of the immediate load bearing
capacity of some commonly used iliac bone grafts. Acta Orthop Scand 1971;
42:482–90.
7. Wittenberg RH, Moeller J, Shea M, et al. Compressive strength of autolo-
gous and allogenous bone grafts for thoracolumbar and cervical spine fusion.
Spine 1990;15:1073– 8.
8. An HS, Xu R, Lim T-H, et al. Prediction of bone graft strength using dual-
energy radiographic absorptiometry. Spine 1994;19:2358 – 62.
9. Lim T-H, Kwon H, Jeon C-H, et al. Effect of endplate conditions and bone
mineral density on the compressive strength of the graft-endplate interface in
anterior cervical spine fusion. Spine 2001;26:951– 6.
10. Ryan PJ, Blake GM, Herd R, et al. Spine and femur BMD by DEXA in
patients with varying severity spinal osteoporosis. Calcif Tissue Int 1993;52:
263– 8.
11. Truumees E, Herkowitz HN. Cervical Spondylotic Radiculopathy, Myelop-
athy, Instructional Course Lectures. Vol 49. Rosemont, IL: American Acad-
emy of Orthopaedic Surgeons; 2000.
12. Emery SE, Bohlman HH, Bolesta MJ, et al. Anterior cervical discectomy and
arthrodesis for the treatment of cervical spondylotic myelopathy. 2–17 year
follow-up. J Bone Joint Surg Am 1998;80A:941–51.
13. White AA III, Jupiter J, Southwick WO, et al. An experimental study of the
immediate load bearing capacity of three surgical constructions for anterior
spine fusions. Clin Orthop 1973;91:21– 8.
14. Truumees E, Demetropoulos CK, Yang KH, et al. Effects of disc height and
distractive forces on graft compression in an anterior cervical discectomy
model. Spine 2002;27:2441–5.
15. Moroney SP, Schultz AB, Miller JAA. Analysis and measurement of neck
loads. J Orthop Res 1988;6:713–20.
16. Enneking WF, Burchardt H, Puhl JJ, et al. Physical and biological aspects of
repair in dog cortical-bone transplants. J Bone Joint Surg Am 1975;57:237–52.