Regional variation in vertebral bone morphology and its contribution to
vertebral fracture strength
P.A. Hulmea,⁎, S.K. Boydb, S.J. Fergusona
aMEM Research Center, University of Bern, Stauffacherstrasse 78, CH 3014, Bern, Switzerland
bSchulich School of Engineering, University of Calgary, Calgary, Alberta, Canada
Received 12 April 2007; revised 6 August 2007; accepted 7 August 2007
Available online 17 August 2007
Vertebral fractures may result in pain, loss of height, spinal instability, kyphotic deformity and ultimately increased morbidity. Fracture risk can
be estimated by vertebral bone mineral density (BMD). However, vertebral fractures may be better defined by more selective methods that account
Our aim was to quantify regional variations in bone architecture parameters (BAPs) and to assess the degree with which regional variations in
BAPs affect vertebral fracture strength. The influence of disc health and endplate thickness on fracture strength was also determined.
The soft tissue and posterior elements of 20 human functional spine units (FSU) were removed (T9 to L5, mean 74.45±4.25 years). After
micro-CT scanning of the entire FSU, the strength of the specimens was determined using a materials testing system. Specimens were loaded in
compression to failure. BAPs were assessed for 10 regions of the vertebral cancellous bone. Disc health (glycosaminoglycan content of the
nucleus pulposus) was determined using the degree of binding with Alcian Blue.
Vertebrae were not morphologically homogeneous. Posterior regions of the vertebrae had greater bone volume, more connections, reduced
trabecular separation and more plate-like isotropic structures than their corresponding anterior regions. Significant heterogeneity also exists
between posterior superior and inferior regions (BV/TV: posterior superior 12.6±2.8%, inferior 14.6±3%; anterior superior 10.5±2.2%, inferior
10.7±2.4%). Of the two endplates that abutted a common disc, the cranial inferior endplate was thicker (0.44±0.15 mm) than the caudal superior
endplate (0.37±0.13 mm). Our study found good correlations between BV/TV, connective density and yield strength. Fracture risk prediction,
using BV/TV multiplied by the cross sectional area of the endplate, can be improved through regional analysis of the underlying cancellous bone
of the endplate of interest (R20.78) rather than analysis of the entire vertebra (R20.65) or BMD (R20.47). Degenerated discs lack a defined
nucleus. A negative linear relationship between disc health and vertebral strength (R20.70) was observed, likely due to a shift in loading from the
weaker anterior vertebral region to the stronger posterior region and cortical shell.
Our results show the importance of considering regional variations in cancellous BAPs and disc health when assessing fracture risk.
© 2007 Elsevier Inc. All rights reserved.
Keywords: Micro-CT; Vertebral body; Trabecular bone; Microstructural properties; Regional variation
Osteoporosis isestimatedtoafflict200millionwomen world-
wide . In the US alone, osteoporotic fractures are estimated to
affect 24 million individuals. 1.5 million new fractures, nearly
half of which are vertebral (700,000), are reported each year,
outnumbering fractures of the hip and ankle combined [2–5].
Vertebral fracture may result in pain at the fracture site, loss of
height due to vertebral collapse, spinal instability and in many
cases a kyphotic deformity . Chronic pain and kyphotic
deformity may lead to depression, decreased appetite (leading to
and a reduction in the quality of life, the ultimate result being a
significant increase in morbidity [7–10]. The World Health
Organization defines osteoporosis as a bone mineral density
(BMD)ofmorethan2.5standarddeviations belowthe meanofa
young healthy reference population of the same gender. How-
ever, BMD only partially determines fracture risk . Various
investigators have found that bone quality, and hence fracture
Bone 41 (2007) 946–957
⁎Corresponding author. Fax: +41 31 631 5960.
E-mail address: Paul.Hulme@MEMcenter.unibe.ch (P.A. Hulme).
8756-3282/$ - see front matter © 2007 Elsevier Inc. All rights reserved.
risk, is independent of BMD, as determined through dual energy
X-ray absorptiometry (DEXA), and have suggested a role for
micro-architecture, turnover, damage accumulation and mineral-
ization [12–17]. DEXA itself does not account for regional
structures that do not add to the mechanical strength of the
vertebra, including posterior elements and osteophytes .
Furthermore, osteoporosis drug treatment strategies have shown
poor correlations between BMD and the risk of vertebral fracture
[19–24]. Hence, vertebral fracture risk may be better defined by
more selective methods.
Regional vertebral morphology has been examined using
histological methods  or using micro-CT measurement of
bone cores [25,26], but few studies consider the analysis of the
vertebral body as a whole . Bone volume for vertebrae has
been determined to be between 6.5% and 16% [11,15,25,28].
Observed age-related architectural changes to the cancellous
bone include: a decrease in bone volume compared with the
total vertebra volume (BV/TV), a shift from plate-like trabe-
culae to more rod-like structures (SMI), a decrease in connec-
tivity density (Conn.D), an increase in orientation of trabeculae
along the axis of principle loading (DA), an increase in trabe-
cular separation (Tb.Sp) and a corresponding decrease in trabe-
culae number (Tb.N) [11,29–31]. Trabecular thickness (Tb.Th)
has been reported to increase or decrease with age [11,31]. An
increase in trabecular thickness has been explained as either
adaptive remodeling of the remaining vertical trabeculae or
removal of the thinner struts resulting in an increase in mean
trabecular thickness [15,26,32]. The etiology of these structural
changes remains unclear, whether it is excessive osteoclastic
resorption or incomplete osteoblast activity resulting in perfora-
tion and thinning of trabecular elements [31,33,34]. However,
the net result is a reduction in vertebral fracture strength.
The aim of this study was to describe the regional morpho-
logy of the elderly human thoracolumbar vertebra using micro-
CTand to relate its derived bone architecture parameters (BAP)
to vertebral failure. The micro-CT gantry allows for entire
functional spine units (FSU) to be scanned non-destructively,
allowing subsequent material tests to be performed using phy-
siologically relevant loading configurations. We hypothesize
that the bone architecture parameters (BAP) of the less dense
anterior and central vertebral body regions will be a better
predictor of failure than those of the whole vertebra or DEXA
measurements [25,28]. We extended our regional analysis to
determine if BAPs can help explain the reported preferential
failure of the superior endplate . Additionally, we consi-
dered the role of disc health on endplate thickness, the under-
lying trabecular bone and vertebral failure.
The surrounding soft tissue and posterior elements of 22 osteoporotic ca-
daveric human functional spine units (FSU) were removed (average 74.45±
4.25 years). Vertebrae were grouped as follows: one T9–T10, three T11–T12,
five T12–L1, five L1–L2, four L2–L3, two L3–L4 and two L4–L5. Previous
investigators have found little difference between male and female bone mor-
phology and no major difference in BV/TV between T9 and L5 (1% T9 to L5,
0.5% T11 to L5). Since morphologic changes due to the aforementioned con-
founding factors are minimal, we pooled all the vertebrae together for analysis
[15,26,36]. Our study population consisted of vertebrae that were used in a
preceding experiment that involved submaximal uniaxial compressive loading
of the FSU. Six of 22 specimens had polymethylmethacrylate (PMMA, Verte-
cem, Synthes, Switzerland) present in the caudal vertebra. Morphology was
assessed prior to cement injection. Bone mineral density (BMD) was assessed
using dual energy x-ray absorptiometry (DEXA) (Discovery C, Hologic, Bed-
Impressions were made of the cranial and caudal endplates in semi-cured
bone cement (Sulfix, Sulzer Orthopaedics Ltd). A jig ensured that the two end
caps were parallel to each other. The molded end caps only extended to the
cortical rim, thereby not adding any structural strength to the FSU. Specimens
were stored in air evacuated polyethylene bags at −20 °C until testing.
Micro-CT scanning and determination of morphology
All measurements were performed on a microcomputed tomography system
(XtremeCT, Scanco Medical AG, Bassersdorf, Switzerland). Scans ranged from
770 to 990 slices with a nominal isotropic resolution of 82 μm (field of view
125 mm, 1536×1536 pixels, integration time 200 ms). Scans were performed in
air, before mechanical testing. Total scan time per specimen approached
30 minutes. Regions of interest (ROI) were identified for trabecular bone and
endplate thickness analysis. The region for trabecular bone analysis included the
cancellous bone from just below the cranial endplate to just above the caudal
endplate, separated into 10 regions (Fig. 1). ROI were filtered using a Laplace–
Hamming filter and segmented using a global threshold. Bone architecture
parameters (BAP) analyzed included (Image Processing Language, v4.29d,
Scanco Medical AG, Bassersdorf, Switzerland): bone volume ratio (BV/TV),
connectivity density (Conn.D), structure model index (SMI), trabecular number
(Tb.N), trabecular thickness (Tb.Th), trabecular separation (Tb.Sp) and degree
of anisotropy (DA) (Table 1). Trabecular thickness, number and separation were
based on direct measures using a distance transformation and not using trabe-
cular model assumptions .
Endplate thickness was determined using a region of interest (ROI) that
to reduce noise. The segmentation threshold was chosen to preserve as much of
Fig. 1. Regions of the vertebra for which bone architecture parameters were
determined. Regions were composed of superior (S), inferior (I), anterior (A),
posterior (P), left (L), right (R) and central (CE) regions.
947 P.A. Hulme et al. / Bone 41 (2007) 946–957
the interconnecting trabeculae as possible without artificially thickening the
endplate or any of the trabeculae. Endplate thickness was determined by
identification of the upper and lower surface of the endplate (custom programs
written in C; Visual Studio 6.0, Microsoft; and Matlab, MathWorks Inc., Natick,
MA). Exploiting the difference in area between trabeculae and trabecular
separation, a 2D median filter was used to remove trabeculae that mayhave been
included in the determination of the lower endplate surface. Endplate thickness
endplate was separated into the ring apophysis and the central endplate region.
The central endplate region was determined by visual inspection of the endplate
height map (the central endplate region is lower than the ring apophysis). User
selected points defined the transition between the ring apophysis and the central
Inc., Natick, MA). The ring apophysis and central endplate region were further
separated into anterior and posterior regions.
Prior to mechanical testing, samples were defrosted overnight in a 0.15-M
NaCl bath at 5 °C. Three hours prior to mechanical testing, the samples were
removed from the refrigerator and allowed to reach room temperature. Testing
was performed on an MTS 858 Mini Bionix II servohydraulic testing machine
(MTS, Eden Prairie, MN, USA). The samples were kept moist with saline
soaked gauze throughout the test. Loading was performed in a similar manner to
our previous FSU mechanical testing procedure . The FSUs were loaded to
50 N and cycled for 600 cycles between 50 and 450 N at 1 Hz for pre-
conditioning.Immediately followingcyclic loading, the sample was compressed
under displacement control at 0.5 mm/s to a total compression of 10 mm. During
compression, load and displacement data were recorded at 100 Hz. Ante-
roposterior and lateral single section scans using micro-CTwere performed after
mechanical testing of the failed FSU to determine the location of the fracture. It
shouldbe notedthat 9 of 22 analyzedspecimens failed duringthe loadingstep of
the preceding experimental series in which specimens were also tested under
axial compression but at a slower loading rate, 0.004±0.003 mm/s. The reduced
loading rate is not thought to have altered the experimental outcome. Ochiva
et al.  found no difference in endplate fracture load between low (10 mm/s)
and high loading rates (2500 mm/s), thus differences between the loading rate of
0.5 mm/s and the lower loading rate used in the previous experimental series
should be negligible.
Disc health assessment
A decrease in the proteoglycan content of the nucleus pulposus has been
associated with disc degeneration [40–43]. To determine proteoglycan content
of the nucleus pulposus (NP), samples were removed from the specimen,
weighed (wet weight) and lyophilized (dry weight). Samples were then digested
with papain (Sigma) in a cysteine buffer (L-cysteine hydrochloride, Fluka) .
Proteoglycan content was determined using the degree of binding with Alcian
Blue (Alcian Blue 8 GX, Fluka) and reported as the ratio of the weight of the
glycosaminoglycan (GAG) side chains divided by the weight of the dry tissue
sample . GAG content was normalized by dry tissue weight since the effect
of mechanical testing, scanning and NP sample removal on disc hydration may
not have been consistent between samples.
Regional variations in bone architecture parameters and endplate thickness
were explored using a one-way ANOVA (Statistica, StatSoft, Tulsa). Newman–
Keuls post hoc analysis was used to identify differences between trabecular
regions. A significance value of p=0.05 was defined. Relationships between
bone quality, endplate thickness, disc health (GAG content) and mechanical
strength; endplate thickness, disc health and bone quality; and disc health and
endplate thickness were assessed using linear regression. The quality of the fit is
reported as the correlation coefficient R2.
Two specimens were removed from the analysis due to an
imaging error on one and the presence of extensive vertebral
sclerosis on the other. The remaining specimens had an average
age of 74.2±4.3 years and were grouped as follows: one T9–
T10, three T11–T12, five T12–L1, four L1–L2, four L2–L3,
one L3–L4 and two L4–L5. The average BAPs for the sample
population (n=20) are given in Table 2. The mean water content
of the nucleus pulposus of the discs was 70±3% and the mean
GAG content was 0.17±0.06 (mg GAG/mg dry tissue, n=20).
Regional morphological variations—trabecular bone
The posterior regions of the vertebrae had greater bone
volume, more connections, more trabeculae, reduced trabecular
separation and more plate-like isotropic structures than their
corresponding anterior regions (Fig. 2). Vertical inhomogeneity
was observed only in the posterior regions. The posterior in-
ferior region had higher bone volume ratio, greater connectivity
density and a greater number of trabeculae, which were more
plate-like in structure and isotropic than its corresponding su-
perior region. It should be noted that there was no statistical
differences noted between any regions using a direct 3D mea-
sure of trabecular thickness . Semi-derived trabecular thick-
ness assessment  resulted in lower thickness measurements
Bone architectural parameters for whole vertebrae
Failure load (N)
Explanation of bone architecture parameters
BV/TV Ratio or % Proportion or percentage of the volume of
interest that is bone
Connectivity density: Indicates the number of
connections normalized by the volume of interest
Structural model index: a quantification of the
prevalent type of trabecular element. 0 for
parallel plate, 3 for cylindrical rods.
Trabecular number: the number of trabeculae
encountered per unit length
Trabecular separation: the average
separation between trabeculae (marrow space)
Degree of anisotropy: a measure of preferential
alignment of the trabeculae along a directional
axis. (1=isotropic, N1 anisotropic)
948P.A. Hulme et al. / Bone 41 (2007) 946–957
(total: 0.139±0.022 mm), less variability but no change in
overall regional distribution. Regions of low trabecular thick-
ness include anterior and central regions.
Regional morphological variations—endplate
The endplate was thinnest in the center and the anterior of the
central endplate region and greatest in the posterior ring apo-
physis (Fig. 3). Differences in endplate thickness were noted
between endplates abutting a common disc. The mean thickness
of the cranial inferior endplate within the central region was
0.44±0.15 mm (max 0.87 mm, min 0.27 mm, n=19) while the
caudal superior endplate thickness was 0.37±0.13 mm (max
0.70 mm, min 0.19 mm, n=19). The inferior cranial endplate
was thicker than the superior caudal endplate for all regions
except the anterior ring apophysis. However, differences were
only statistically significant for the posterior central endplate
region (Fig. 3).
Fig. 2. Regional variation in bone architecture parameters. Data are displayed as the mean bone architecture parameter±95% C.I. (1) Significant difference between
transverseregions(i.e. anteriorandposterior)of the samevertical region(i.e.superiororinferior);(2) significantdifferencebetweenvertical regionsof sametransverse
949P.A. Hulme et al. / Bone 41 (2007) 946–957
There were large variations in endplate thickness between
specimens, dependent upon the presence of sclerotic bone
formation. The molded endcaps, created to facilitate mechan-
ics testing, were present during the micro-CT scans. Unfor-
tunately, the contrast between the molds and the endplate was
not sufficient to resolve the two components; thus, analysis
of the superior cranial and inferior caudal endplate was not
A positive correlation was noted between the central region's
endplate thickness and the BV/TVof the underlying trabecular
bone (R20.44, pb0.001).
Bone morphology parameters that best predict vertebral failure
Pearson's correlation coefficients between regional BAPs
Correlations were improved through consideration of vertebral
geometry (BAP×CSA). Cross-sectional area (CSA) of the
endplates was determined from the micro-CTscans, after digital
removal of the osteophytes. While the BAPs derived from the
analysis of an entire vertebra had good correlations with failure
strength, it was not until regional variations were taken into
consideration that significant improvements were noted over
traditional BMD measurements (Figs. 5, 6). Regional measures
of connectivity density and BV/TV best predicted vertebral
failure load (BV/TV R2=0.78 Fig. 7, Conn.D R2=0.70
Table 3). The best correlations between BAPs and fracture
strength were obtained when the trabecular zone below the
fractured endplate was analyzed (Fig. 5), confirming the role of
regional BAPs in fracture prediction. There was also significant
transverse variability in the ability of BAPs to predict vertebral
strength. With few exceptions the posterior region correlated
best with fracture load (Fig. 6), possibly due to the loading
protocol and the effects of disc health [46,47].
Preferential failure of the superior endplate may be explained
through morphological analysis of the vertebrae which abut a
common disc [35,48,49]. During our experimental series, 13 of
14 FSUs (without cement present in the caudal vertebra) expe-
rienced caudal superior endplate fractures. Cranial inferior end-
plate fractures occurred in those FSUs whose caudal vertebrae
were augmented with PMMA. The superior endplate caudal to a
only the posterior region's differences reached statistical signi-
ficance(Fig. 3). However, since only a moderate correlation was
determined between endplate thickness and vertebral yield
is the sole cause for preferential superior endplate failure. The
While no difference could be detected between the underlying
anterior trabecular bone region of the cranial inferior and caudal
superior endplate, the cranial posterior trabecular region does
density, more plate like elements and increased trabecular num-
ber (p=0.04, 0.01, 0.05, 0.01, respectively, n=20).
Disc health (GAG content of the NP) and morphology
There was a weak negative correlation noted between BV/
TV beneath the endplate and disc health (R20.19, p=0.006). A
negative correlation was found between disc health and the
central endplate thickness (R20.33, pb0.001). Disc health was
assessed by determining the ratio of mg GAG to mg dry tissue.
The effect of disc degeneration on vertebral fracture strength
Disc health had a significant effect on vertebral failure load
(Fig. 8). Disc health had a negative linear relationship with
vertebral failure strength (R20.70, pb0.001).
Fig. 3. Regional thickness of the caudal and cranial endplate. Insert illustrates a typical endplate thickness map, the white boundary indicates the central endplate
region.Darkregionsarethinnerthanlighterregions(n=19 inferior,19 superior endplates;RA:ring apophysis;CER:centralendplateregion;⁎pb0.05betweencranial
inferior and caudal superior endplates;#pb0.05 between regions on either the cranial or caudal endplate). Data are displayed as the mean±1 SD.
950 P.A. Hulme et al. / Bone 41 (2007) 946–957
Vertebrae are not morphologically homogeneous. While
vertical homogeneity was observed for the anterior trabecular
regions, there is considerable posterior vertical and transverse
inhomogeneity, confirming previous reports. The presented
methodology differs from previous investigations in that it is not
limited to histological analysis of specific regions or the
analysis of regional bone plugs, thus allowing non-invasive
morphology assessment of the entire vertebral body, followed
by mechanical analysis [11,17,25,26,28,36,50]. This method-
ology allows the direct assessment of the relation between
structure and function. Banse et al.  reported a 25% and
20% difference in bone volume between the superior and
inferior posterior regions for thoracolumbar and lumbar
vertebrae, respectively. Our results are similar, with a combined
Pearson's correlation coefficients relating bone architecture parameters and vertebral failure strength
Bone architecture parameterRegion closes to fracture Region furthest from FxSuperior regionInferior region
BAP−Load BAP−Stress BAP×CSA−LoadBAP×CSA−Load BAP×CSA−Load
BMD0.34 0.350.49 0.490.49 0.49
Average anterior and posterior
Average anterior and posterior
Average anterior and posterior
Average anterior and posterior
Average anterior and posterior
Average anterior and posterior
Average anterior and posterior
951P.A. Hulme et al. / Bone 41 (2007) 946–957
thoracolumbar and lumbar difference of 16%. The ratio of the
average anterior and posterior ash density (1.24), reported by
Nepper-Rasmussen and Mosekilde , agrees with the trans-
verse variation of BV/TV found in the current study (1.28, ratio
of mean posterior BV/TV to mean anterior BV/TV). The re-
vertebral fracture location. Low energy compression fractures
can occur in the anterior or posterior superior regions of the
vertebra, but rarely the inferior posterior region [28,51,52]. One
inferior posterior region, is that the cortical shell of the superior
posterior regionis re-enforced by the posterior elements, relying
less on the trabeculae for strength (Fig. 9). However, the inferior
posterior region of the cortical shell does not have re-enforce-
ment, lacking any posterior elements and thus must rely on the
trabecular core for strength .
The increased BV/TV of the posterior inferior vertebral re-
gion, and to a lesser extent a thickening of the posterior end-
plate, likely strengthens the inferior endplate resulting in a
preferential failure of the superior endplate. Our results show
how vertebral morphology can influence function. In the study
by Grant et al. , only the posterior inferior endplate was
found to be significantly stronger than the superior endplate.
Similarly, in the present study the greatest increase in bone
quality and endplate thickness was noted in the inferior poste-
rior region (Fig. 3).
The relationship between vertebral strength and structure is
complex, involving contributions from the endplate, cortical
shell, trabeculae and intervertebral disc health . The present
study found good correlations between BV/TVand Conn.D and
yield strength (Fig. 5). There was greater variability in Conn.D
and to a lesser extent BV/TV between specimens, compared
with the other BAP parameters. A high standard deviation could
indicate BAPs that are sensitive to remodeling changes (hence
useful for fracture prediction), or a reflection of measurement
inaccuracies. However, both BV/TV and Conn.D correlated
strongly with fracture strength; thus, the first hypothesis is more
probable. It should be noted that a small variability in BAP does
not necessarily reflect poor ability to predict fracture strength,
only that appropriate equipment is required to resolve the dif-
ferences between samples.
its importance in the maintenance of bone strength (Table 3). As
bone volume decreases, there is a corresponding decrease in
connectivity density (Fig. 10), possibly due to the loss of small
interconnecting trabeculae with small initial diameter . The
ratio of vertical trabeculae to horizontal, while almost 2:1 for
Fig. 4. Correlation of bone architecture parameters evaluated for the vertebra as
a whole with measured failure strength (n=20, SMI is divided by CSA, all
Fig. 5. Correlation between regional BAP×CSA and failure load for various vertebral regions (SMI is divided by CSA,§pN0.05).
952 P.A. Hulme et al. / Bone 41 (2007) 946–957
young individuals, increases with age [32,55]. Thus, trabecular
perforation and subsequent removal occurs preferentially for
an increased susceptibility to buckling, a recognized primary
mode of failure for cancellous bone [53,55–57].
A linear regression model was used to determine the corre-
lation between bone architecture and failure load and stress.
Although the use of a power model is more biologically correct,
Cody et al.  and Ebbesen et al.  found minimal im-
provement using the power relationship. The inclusion of ex-
variation in bone volume, a linear model is valid .
Correlation between failure and bone architecture parameters
(BAP) can be improved if, as other authors have suggested, load
is correlated to BAP×CSA [60–62]. In addition to multiplying
BAP by CSA, we tried other fracture prediction schemes .
the standard deviation of regional BAP or the relative density
index (ratio of the maximum and minimum regional BAP), two
schemes postulated to reflect bone remodeling changes .
The correlation of BAPs with failure stress or load differed
according to the region analyzed. We did not find improved
correlations between BAPs and yield for the anterior or central
region as was hypothesized (Table 3, Fig. 6), possibly due to the
loading protocol which preferentially loads the posterior regions
. For both the normal and degenerated FSU, the majority of
the load is carried by the posterior vertebral body during upright
postures . Cody et al.  found that bone mineral density
had the highest correlation with failure load in the inferior
region. This region had moderate correlation between BV/TV
and fracture load in the present study but correlation was higher
when analyzing the superior region or the entire vertebra .
One would expect that the weakest zone (superior half of the
vertebra) would dictate at what load the vertebra would fail.
While good correlations were found between the BAPs of the
Fig. 6. Correlation of BAP×CSA and vertebral failure load for various transverse regions (SMI is divided by CSA, all correlations pb0.05).
Fig. 7. A linear relationship exists between BV/TV×CSA and vertebral fracture strength. The subchondral bone below the fractured endplate was analyzed (mean of
the anterior and posterior region).
953P.A. Hulme et al. / Bone 41 (2007) 946–957
superior region and vertebral strength, there was no improve-
ment over using analysis of the entire vertebra, likely due to the
two fracture modalities that were created (caudal superior
endplate fracture vs. cranial mixed fracture mode). Vertebral
fracture strength could not be explained through analysis of one
specific region, instead BAPs from the region in which the
fracture occurred (Fig. 5) correlated best with vertebral fracture
strength. This result highlights the importance of regional
assessments for fracture risk.
The loading protocol of this study was limited to uniaxial
compression, resulting in ostensibly endplate fractures with
some cortical involvement. However, most elderly fractures that
occur are wedge fractures, the result of applied compression and
bending moment to induce flexion [63,64].One would expect in
this situation that the importance of BAPs of the anterior ver-
tebral body region might predict fracture strength more effec-
tively. Due to the bore size of the micro-CT, posterior elements
were removed from the specimens; therefore, the contribution
of the neural arch to load bearing was not considered. The
neural arch only carries a minor fraction of the load for an FSU
with a normal disc under axial compression. However, the
presence of a degenerative disc shifts a considerable portion of
the load to the neural arch [46,65]. Thus, our experimental
construct represents a worst case scenario, when the entire load
fraction is carried by the vertebral body. The lack of posterior
elements may also explain the low failure loads recorded.
Changes in loading pattern from the in vivo situation (i.e.,
elimination of the neural arch) may shift regional trabecular
loading anteriorly resulting in early failure of the vertebra.
Previous quantitative computed tomography (QCT) and
DEXA studies have found correlations between vertebral
strength and the product of QCT bone minimal density (or
DEXA) and cross-sectional area between R2=0.45 to 0.88 for
multiple vertebral segment constructs [58,60,66–68] and 0.26 to
0.82 for single vertebra tests [61,62,69–71]. Given our study
population and the relatively unrestrictive inclusion criteria, the
correlation between micro-CT derived BAP×CSA and failure
load was good (R2=0.78). We hypothesize that strict inclusion
have shownonly moderatecorrelationbetweenBMD×CSAand
Fig. 8. Correlation of disc health (GAG content of the nucleus pulposus) with vertebral failure strength.
Fig. 9. Vertical morphological heterogeneity is evident in this sagittal slice through a vertebral specimen. Bone volume ratio and connectivity density are higher in the
posterior inferior region. Sections (a) and (b) illustrate a transverse slice near the superior and inferior endplates, note both the presence of transverse heterogeneity
within the slicesand vertical heterogeneitybetween the posteriorregions of the slices. Boxes highlightthe regions from whichthe slices wereconstructed (sagittalslice
is 1.64 mm thick, transverse slices are 2.46 mm thick). Arrows indicate osteophytes.
954 P.A. Hulme et al. / Bone 41 (2007) 946–957
load (R2=0.47). AP DEXA is a general measurement of bone
quality, including contributions from the cancellous bone, cor-
tical shell and posterior elements, and is influenced by sclerotic
growth andosteophytes. The experimental protocol required that
and micro-CT scanning. However, remnants still remained to a
varying degree on the FSU, which would affect the BMD, in-
creasing values for those with greater remnants without a corres-
ponding increase in mechanical strength. Some specimens also
had considerable osteophytes (evident on Fig. 9a), which do not
addto themechanicalstrength of thevertebrae,butincrease their
of allowing the volume of analysis to be specified, allowing the
user to remove any anomalous regions.
We did not consider the contribution of the cortical shell to
vertebral fracture strength. The percentage of the load carried by
the cancellous bone and cortical shell varies dependent upon the
distance from the endplates [27,47,72–74]. Near the endplates,
the majority of the load is carried by the cancellous bone, which
changes to about equal sharing between the cortical shell and
cancellous bone at the midtransverse plane [27,47,72–74]. All
of the specimens in this experimental series had endplate
fractures, a region in which most of the load is carried by the
cancellous core [27,47,72–75]. Thus, only BAPs for the can-
cellous core were correlated to fracture strength. However, this
is not to say that the cortical shell has no role in fracture
prediction. An improvement in fracture prediction was noted by
Andresen et al.  after considering the contribution of both
the BMD of the cancellous bone and cortical shell. The impor-
tance of the cortical shell for structural strength, especially for
osteoporotic vertebrae, has been highlighted by various other
investigators [27,74,76,77]. Future studies should consider the
combined role of cortical shell and cancellous bone in fracture
We found a strong negative correlation between disc health
(GAG content) and endplate failure load (Fig. 8). A negative
correlation also exists between disc health and endplate thick-
ness and BV/TV. While specimens with poor disc health typi-
cally had increased BV/TVand endplate thickness, the effect of
disc health on bone quality alone cannot explain the correlation
between disc health and fracture strength. A shift in load trans-
fer between vertebral bodies also occurs for degenerative discs.
Degenerative discs lack a defined nucleus and under axial
compressive loads typically have lower stresses in the anterior
half of the disc . Load shift towards the periphery of the
vertebral body and the posterior elements decreases the amount
of bone at risk of fracture. A shift in loading from the weaker
anterior vertebral region to the stronger posterior region results
in higher vertebral fracture strength [46,47].
A limitation to the micro-CT analysis is that the measure-
ments were performed using a nominal resolution of 82 μm,
which is close to the mean thickness of human trabeculae, thus
partial volume effects may result in a measurement bias. How-
ever, to reasonably determine bone architecture, resolutions
below 100 μm have been shown to be acceptable [78,79]. The
performance of the micro-CT scanner used in this study
(XtremeCT) has been assessed versus the gold standard of a
high-resolution micro-CT scanner (19 μm) . It was found
that bone architecture parameters BV/TV, Conn.D, Tb.N, Tb.Sp
and DA performed well. Although BV/TV was over estimated
by 28%, it was a significant improvement over the 60% re-
ported by Banse et al.  using different equipment. The
resolution of the scanner can result in measurement inaccura-
cies for the parameters, Tb.Th and SMI ; however, it should
be noted that in the present study, we were not constrained to in
vivo scanning protocols, thus integration time could be in-
creased. This yields improved image contrast and likely im-
proved determination of bone architecture parameters over the
previously reported in vivo protocol-based accuracy and pre-
cision assessments [79,80].
Our research has demonstrated the relation between whole
vertebra micro-architectural parameters and fracture strength
using micro-CT and identified regional variation in trabecular
morphology. In addition we have identified disc health as an
Fig. 10. A linear relationship exists between bone volume ratio and the connectivity density—as the bone volume ratio decreases, connections are lost between
955P.A. Hulme et al. / Bone 41 (2007) 946–957
important contributor to vertebral fracture prediction. Micro-CT
currently cannot be used clinically to directly assess bone qua-
lity of vertebral bodies and to date the relation between peri-
pheral X-ray-based analysis and vertebral failure strength lacks
adequate correlation [66,70]. In vivo assessment using lower
resolution clinical CT scanners permit reasonable estimates of
BV/TV but may not have the resolution required to estimate
Conn.D, the two parameters that this study identifies as best
correlating with fracture strength . However, an under-
standing of regional morphology may be used to focus DEXA
or clinical CT analysis to specific trabecular zones to further
improve fracture prediction for specific fracture modalities.
Funding for this research was provided by AOSpine (Grant #
SRN 02/105). The authors wish to thank Josh MacNeil for his
assistance with micro-CT data acquisition and analysis, Jennifer
Vuong for lyophilizing the NP samples and Ladina Ettinger for
her assistance with the GAG analysis.
 Iqbal MM. Osteoporosis: epidemiology, diagnosis, and treatment. South
Med J 2000;93:2–18.
 Cooper C, Atkinson EJ, Jacobsen SJ, O'Fallon WM, Melton III LJ. Popu-
lation-based study of survival after osteoporotic fractures. Am J Epidemiol
 Truumees E, Hilibrand A, Vaccaro AR. Percutaneous vertebral augmen-
tation. Spine J 2004;4:218–29.
 Zoarski GH, Snow P, Olan WJ, Stallmeyer MJ, Dick BW, Hebel JR, et al.
Percutaneous vertebroplasty for osteoporotic compression fractures: quan-
titative prospective evaluation of long-term outcomes. J Vasc Interv Radiol
 Riggs BL, Melton III LJ. The worldwide problem of osteoporosis: insights
afforded by epidemiology. Bone 1995;17:505S–11S.
 Silverman SL. The clinical consequences of vertebral compression frac-
ture. Bone 1992;13(Suppl 2):S27–31.
 Cooper C, Atkinson EJ, O'Fallon WM, Melton III LJ. Incidence of clini-
cally diagnosed vertebral fractures: a population-based study in Rochester,
Minnesota, 1985–1989. J Bone Miner Res 1992;7:221–7.
 Coumans JV, Reinhardt MK, Lieberman IH. Kyphoplasty for vertebral
compression fractures: 1-year clinical outcomes from a prospective study.
J Neurosurg 2003;99:44–50.
 Lyles KW, Gold DT, Shipp KM, Pieper CF, Martinez S, Mulhausen PL.
Association of osteoporotic vertebral compression fractures with impaired
functional status. Am J Med 1993;94:595–601.
 Evans AJ, Jensen ME, Kip KE, DeNardo AJ, Lawler GJ, Negin GA, et al.
Vertebral compression fractures: pain reduction and improvement in func-
tional mobility after percutaneous polymethylmethacrylate vertebroplasty
retrospective report of 245 cases. Radiology 2003;226:366–72.
 Hordon LD, Raisi M, Aaron JE, Paxton SK, Beneton M, Kanis JA.
Trabecular architecture in women and men of similar bone mass with and
without vertebral fracture: I. Two-dimensional histology. Bone 2000;27:
 Hui SL, Slemenda CW, Johnston Jr CC. Age and bone mass as predictors
of fracture in a prospective study. J Clin Invest 1988;81:1804–9.
 Burr DB, Forwood MR, Fyhrie DP, Martin RB, Schaffler MB, Turner CH.
Bone microdamage and skeletal fragility in osteoporotic and stress
fractures. J Bone Miner Res 1997;12:6–15.
 Ciarelli TE, Fyhrie DP, Schaffler MB, Goldstein SA. Variations in three-
dimensional cancellous bone architecture of the proximal femur in female
hip fractures and in controls. J Bone Miner Res 2000;15:32–40.
 Stauber M, Müller R. Age-related changes in trabecular bone micro-
structures:globalandlocal morphometry. Osteoporos Int 2006;17:616–26.
 Homminga J, McCreadie BR, Ciarelli TE, Weinans H, Goldstein SA,
Huiskes R. Cancellous bone mechanical properties from normals and
patients with hip fractures differ on the structure level, not on the bone hard
tissue level. Bone 2002;30:759–64.
 Legrand E, Chappard D, Pascaretti C, Duquenne M, Krebs S, Rohmer V,
et al. Trabecular bone microarchitecture, bone mineral density, and verte-
bral fractures in male osteoporosis. J Bone Miner Res 2000;15:13–9.
 Mosekilde L. Age-related changes in bone mass, structure, and strength-
effects of loading. Z Rheumatol 2000;59(Suppl 1):1–9.
 Eastell R, Barton I, Hannon RA, Chines A, Garnero P, Delmas PD.
Relationship of early changes in bone resorption to the reduction in
fracture risk with risedronate. J Bone Miner Res 2003;18:1051–6.
 Cummings SR, Karpf DB, Harris F, Genant HK, Ensrud K, LaCroix AZ,
et al. Improvement in spine bone density and reduction in risk of vertebral
fractures during treatment with antiresorptive drugs. Am J Med 2002;112:
 Sarkar S, Mitlak BH, Wong M, Stock JL, Black DM, Harper KD.
Relationships between bone mineral density and incident vertebral fracture
risk with raloxifene therapy. J Bone Miner Res 2002;17:1–10.
 Watts NB, Cooper C, Lindsay R, Eastell R, Manhart MD, Barton IP, et al.
Relationship between changes in bone mineral density and vertebral
fracture risk associated with risedronate: greater increases in bone mineral
density do not relate to greater decreases in fracture risk. J Clin Densitom
 Watts NB, Geusens P, Barton IP, Felsenberg D. Relationship between
changes in BMD and nonvertebral fracture incidence associated with
risedronate: reduction in risk of nonvertebral fracture is not related to
change in BMD. J Bone Miner Res 2005;20:2097–104.
 Sarkar S, Reginster JY, Crans GG, Diez-Perez A, Pinette KV, Delmas PD.
Relationship between changes in biochemical markers of bone turnover
and BMD to predict vertebral fracture risk. J Bone Miner Res 2004;19:
 Gong H, Zhang M, Qin L, Lee KK, Guo X, Shi SQ. Regional variations in
microstructural properties of vertebral trabeculae with structural groups.
 Stauber M, Muller R. Age-related changes in trabecular bone micro-
structures:globalandlocal morphometry. Osteoporos Int 2006;17:616–26.
 Eswaran SK, Gupta A, Adams MF, Keaveny TM. Cortical and trabecular
load sharing in the human vertebral body. J Bone Miner Res 2006;21:
 Banse X, Devogelaer JP, Munting E, Delloye C, Cornu O, Grynpas M.
Inhomogeneity of human vertebral cancellous bone: systematic density
and structure patterns inside the vertebral body. Bone 2001;28:563–71.
 Hildebrand T, Ruegsegger P. Quantification of bone microarchitecturewith
the Structure Model Index. Comput Methods Biomech Biomed Engin
 Gong H, Zhang M, Yeung HY, Qin L. Regional variations in microstruc-
tural properties of vertebral trabeculae with aging. J Bone Miner Metab
 Parfitt AM, Mathews CH, Villanueva AR, Kleerekoper M, Frame B, Rao
DS. Relationships between surface, volume, and thickness of iliac trabe-
cular bone in aging and in osteoporosis. Implications for the micro-
anatomic and cellular mechanisms of bone loss. J Clin Invest 1983;72:
 Thomsen JS, Ebbesen EN, Mosekilde LI. Age-related differences between
thinning of horizontal and vertical trabeculae in human lumbar bone as
assessed by a new computerized method. Bone 2002;31:136–42.
 Kleerekoper M, Villanueva AR, Stanciu J, Rao DS, Parfitt AM. The role of
three-dimensional trabecular microstructure in the pathogenesis of ver-
tebral compression fractures. Calcif Tissue Int 1985;37:594–7.
 Mosekilde L. Sex differences in age-related loss of vertebral trabecu-
lar bone mass and structure-biomechanical consequences. Bone 1989;10:
 Roberts S, McCall IW, Menage J, Haddaway MJ, Eisenstein SM. Does the
thickness of the vertebral subchondral bone reflect the composition of the
intervertebral disc? Eur Spine J 1997;6:385–9.
956P.A. Hulme et al. / Bone 41 (2007) 946–957
 Grote HJ, Amling M, Vogel M, Hahn M, Posl M, Delling G. Intervertebral Download full-text
variation in trabecular microarchitecture throughout the normal spine in
relation to age. Bone 1995;16:301–8.
 Hildebrand T, Rüegsegger P. A new method for the model-independent
assessment of thickness in three-dimensional images. J Microsc 1997;185:
 Berlemann U, Ferguson SJ, Nolte LP, Heini PF. Adjacent vertebral failure
after vertebroplasty. A biomechanical investigation. J Bone Joint Surg Br
 Ochia RS, Tencer AF, Ching RP. Effect of loading rate on endplate and
vertebral body strength in human lumbar vertebrae. J Biomech 2003;36:
 Pearce RH, Grimmer BJ, Adams ME. Degeneration and the chemical
compositionof the human lumbar intervertebral disc. J Orthop Res 1987;5:
 Guiot BH, Fessler RG. Molecular biology of degenerative disc disease.
T1rho-weighted magnetic resonance imaging. Spine 2006;31:1253–7.
 Cs-Szabo G, Ragasa-San Juan D, Turumella V, Masuda K, Thonar EJ, An
HS. Changes in mRNA and protein levels of proteoglycans of the anulus
fibrosus and nucleus pulposus during intervertebral disc degeneration.
A Practical Approach. Washington, D.C.: Oxford University Press; 1994.
 Bjornsson S. Simultaneous preparation and quantitation of proteoglycans
by precipitation with Alcian blue. Anal Biochem 1993;210:282–91.
 Pollintine P, Dolan P, Tobias JH, Adams MA. Intervertebral disc degene-
ration can lead to “stress-shielding” of the anterior vertebral body: a cause
of osteoporotic vertebral fracture? Spine 2004;29:774–82.
 Homminga J, Weinans H, Gowin W, Felsenberg D, Huiskes R. Osteo-
porosis changes the amount of vertebral trabecular bone at risk of fracture
but not the vertebral load distribution. Spine 2001;26:1555–61.
 Grant JP, Oxland TR, Dvorak MF. Mapping the structural properties of the
lumbosacral vertebral endplates. Spine 2001;26:889–96.
 Zhao F, Pollintine P, Hole B, Adams M, Dolan P. Why do vertebral
fractures usually affect the cranial endplate? Proceedings of the Annual
Congress of the International Society for the Study of the Lumbar Spine,
(June 13–17, 2006: Bergen, Norway); 2006.
 Nepper-Rasmussen J, Mosekilde L. Local differences in mineral content in
vertebral trabecular bone measured by dual-energy computed tomography.
Acta Radiol 1989;30:369–71.
 Silva MJ, Keaveny TM, Hayes WC. Computed tomography-based finite
element analysis predicts failure loads and fracture patterns for vertebral
sections. J Orthop Res 1998;16:300–8.
 Magerl F, Aebi M, Gertzbein SD, Harms J, Nazarian S. A comprehensive
classification of thoracicandlumbar injuries.Eur Spine J1994;3:184–201.
 Silva MJ, Gibson LJ. Modeling the mechanical behavior of vertebral
trabecular bone: effects of age-related changes in microstructure. Bone
 Guo XE, Kim CH. Mechanical consequence of trabecular bone loss and its
treatment: a three-dimensional model simulation. Bone 2002;30:404–11.
 McDonnell P, McHugh PE, O'Mahoney D. Vertebral osteoporosis and
trabecular bone quality. Ann Biomed Eng 2007;35:170–89.
 Rubin CD. Emerging concepts in osteoporosis and bone strength. Curr
Med Res Opin 2005;21:1049–56.
 Davison KS, Siminoski K, Adachi JD, Hanley DA, Goltzman D, Hodsman
AB, et al. Bone strength: the whole is greater than the sum of its parts.
Semin Arthritis Rheum 2006;36:22–31.
 Cody DD, Goldstein SA, Flynn MJ, Brown EB. Correlations between
vertebral regional bone mineral density (rBMD) and whole bone fracture
load. Spine 1991;16:146–54.
 Ebbesen EN, Thomsen JS, Beck-Nielsen H, Nepper-Rasmussen HJ,
Mosekilde L. Lumbar vertebral body compressive strength evaluated by
dual-energy X-ray absorptiometry, quantitativecomputedtomography, and
ashing. Bone 1999;25:713–24.
 Biggemann M, Hilweg D, Brinckmann P. Prediction of the compressive
strength of vertebral bodies of the lumbar spine by quantitative computed
tomography. Skeletal Radiol 1988;17:264–9.
 Cheng XG, Nicholson PH, Boonen S, Lowet G, Brys P, Aerssens J, et al.
Prediction of vertebral strength in vitro by spinal bone densitometry and
calcaneal ultrasound. J Bone Miner Res 1997;12:1721–8.
 Singer K, Edmondston S, Day R, Breidahl P, Price R. Prediction of
thoracic and lumbar vertebral body compressive strength: correlations with
bone mineral density and vertebral region. Bone 1995;17:167–74.
 Wasnich RD. Vertebral fracture epidemiology. Bone 1996;18:179S–83S.
 Melton III LJ, Kan SH, Frye MA, Wahner HW, O'Fallon WM, Riggs BL.
Epidemiology of vertebral fractures in women. Am J Epidemiol 1989;129:
 Pollintine P, Przybyla AS, Dolan P, Adams MA. Neural arch load-bearing
in old and degenerated spines. J Biomech 2004;37:197–204.
 Lochmuller EM, Burklein D, Kuhn V, Glaser C, Müller R, Gluer CC, et al.
Mechanical strength of the thoracolumbar spine in the elderly: prediction
from in situ dual-energy X-ray absorptiometry, quantitative computed
tomography (QCT), upper and lower limb peripheral QCT, and quanti-
tative ultrasound. Bone 2002;31:77–84.
 Moro M, Hecker AT, Bouxsein ML, Myers ER. Failure load of thoracic
vertebrae correlates with lumbar bone mineral density measured by DXA.
Calcif Tissue Int 1995;56:206–9.
 Ito M, Ikeda K, Nishiguchi M, Shindo H, Uetani M, Hosoi T, et al. Multi-
detector row CT imaging of vertebral microstructure for evaluation of
fracture risk. J Bone Miner Res 2005;20:1828–36.
 Edmondston SJ, Singer KP, Day RE, Price RI, Breidahl PD. Ex vivo
estimation of thoracolumbar vertebral body compressive strength: the
relative contributions of bone densitometry and vertebral morphometry.
Osteoporos Int 1997;7:142–8.
 Bjarnason K, Hassager C, Svendsen OL, Stang H, Christiansen C. Antero-
posterior and lateral spinal DXA for the assessment of vertebral body
strength: comparison with hip and forearm measurement. Osteoporos Int
 Crawford RP, Cann CE, Keaveny TM. Finite element models predict in
vitro vertebral body compressive strength better than quantitative com-
puted tomography. Bone 2003;33:744–50.
 Homminga J, Van Rietbergen B, Lochmuller EM, Weinans H, Eckstein F,
Huiskes R. The osteoporotic vertebral structure is well adapted to the loads
of daily life, but not to infrequent “error” loads. Bone 2004;34:510–6.
 Silva MJ, Keaveny TM, Hayes WC. Load sharing between the shell and
centrum in the lumbar vertebral body. Spine 1997;22:140–50.
 Cao KD, Grimm MJ, Yang KH. Load sharing within a human lumbar
vertebral body using the finite element method. Spine 2001;26:E253–60.
 Alonso CA, Bermejo AL, Yuste Pena AS, Garcia DL. Epidural anesthesia
for percutaneous vertebroplasty. Rev Esp Anestesiol Reanim 2003;50:
 Andresen R, Werner HJ, Schober HC. Contribution of the cortical shell of
vertebrae to mechanical behaviour of the lumbar vertebrae with impli-
cations for predicting fracture risk. Br J Radiol 1998;71:759–65.
 Ritzel H, Amling M, Posl M, Hahn M, Delling G. The thickness of human
vertebral cortical bone and its changes in aging and osteoporosis: a histo-
morphometric analysis of the complete spinal column from thirty-seven
autopsy specimens. J Bone Miner Res 1997;12:89–95.
 Müller R, Rüegsegger P. Micro-tomographic imaging for the nondestruc-
tive evaluation of trabecular bone architecture. In: Lowet G, editor. Bone
computed tomography for measurement of bone quality. Med Eng Phys
 Boutroy S, Bouxsein ML, Munoz F, Delmas PD. In vivo assessment of
trabecular bone microarchitecture by high-resolution peripheral quantita-
tive computed tomography. J Clin Endocrinol Metab 2005;90:6508–15.
957 P.A. Hulme et al. / Bone 41 (2007) 946–957