3d bone micro archit

8/3/2019 3D Bone Micro Archit

http://slidepdf.com/reader/full/3d-bone-micro-archit 1/6

Side-to-side and within-side variability of 3D bone microarchitecture by conventional

micro-computed tomography of paired iliac crest biopsies

C. Chappard ⁎, A. Marchadier, C.L. Benhamou

INSERM U658, Centre Hospitalier Régional, 1 rue Porte Madeleine, 45032 Orléans, France

a r t i c l e i n f o a b s t r a c t

Article history:

Received 21 November 2007

Revised 24 January 2008

Accepted 16 February 2008

Available online 18 March 2008

Edited by: R. Recker

Bone microarchitecture in osteoporosis can be characterized by examining iliac bone biopsies and

treatment effects assessed by comparing a baseline biopsy from one side to a posttreatment biopsy fromthe other side, a method that assumes limited side-to-side variability. New techniques based on micro-

computed tomography (µCT) provide information on the three-dimensional (3D) microarchitecture of

bone. We used µCT to measure side-to-side and within-side variability of 3D microarchitectural

parameters of trabecular and cortical bone in paired iliac-crest biopsies, one from each side. A Bordier

needle trephine was used to collect biopsies from 30 postmenopausal female cadavers (mean age, 73.7±

10.7 years; range, 55–96 years). Biopsies were chemically defatted then imaged using a desktop µCT

scanner (voxel size, 10.77 µm). Parameters measured in trabecular bone consisted of bone volume/tissue

volume (BV/TV, %), direct trabecular thickness and trabecular spacing (Tb.Th⁎ and Tb.Sp⁎, µm) using the

sphere method, bone surface/bone volume (BS/BV, mm−1), trabecular number (Tb.N, mm−1), structure

model index (SMI), trabecular pattern factor (Tb.Pf), and degree of anisotropy (DA). In cortical bone, we

measured cortical thickness (Cort.Th), porosity (Cort.Porosity), and pore diameter (Po.Dm). For trabecular

bone parameters, reproducibility as assess ed from two µCT acquisitions ranged from 4.1% to 6.9%. To assess

side-to-side variability, we matched the volumes of interest selected in the right and left iliac crests. The

mean difference in absolute individual percent variation (mAbsΔind) between the two sides ranged from

10.8% to 14.8% for all trabecular parameters except Tb.Pf (74%) and SMI (84%). In cortical bone, mAbs Δind

were 11.6% for Po.Dm, 15.1% for Cort.Porosity, and 27.6% for Cort.Th. To assess within-side variability, wedivided the trabecular iliac crest volume into three equal parts, one adjacent to each cortex and one in the

middle. Values of mAbsΔind versus the middle part were ranging from 7.6% for Tb.Sp⁎ to 26.2% for BV/TV.

Thus, within-side variability was similar in magnitude to side-to-side variability. The considerable

differences in robustness across trabecular parameters indicate a need for selecting the most stable

parameters, most notably for longitudinal studies of small numbers of patients. Acquisition by µCT and

image analysis must comply with stringent quality criteria, especially the distance from the cortices must

be standardized.

© 2008 Elsevier Inc. All rights reserved.

Keywords:

Trabecular bone

Cortical bone

Micro-architecture

Micro-CT

Iliac crest

Introduction

The definition of osteoporosis has changed over the years and now

includes alterations in bone quality in addition to bone loss [1]. Bone

microarchitecture is the most extensively studied component of bone

quality. Two-dimensional (2D) histomorphometric analysis of bone

biopsy sections was long considered the reference standard for

assessing trabecular bone microarchitecture [2].

The iliac crest is the most widely used biopsy site in studies of

metabolic bone disease [3]. Examination of iliac crest specimens

provides information on bone microarchitecture, mineralization, and

the effects of treatments. Iliac crest biopsy reliably predicted the

amount of mineralized trabecular bone in the lumbar spine [4].

Compressive strength of iliac trabecular bone correlated closely with

that of vertebral trabecular bone [5]. Bone histomorphometry

measurements obtained from iliac crest biopsies have been used to

characterize microarchitectural changes in patients with or without

vertebral fractures [6], women with corticosteroid-induced osteo-

porosis [7], and men with osteoporosis [8]. Iliac crest biopsies were

performed to compare women with primary hyperparathyroidism,

women with osteoporosis, and healthy women [9]. Histological

studies of iliac crest biopsies have shed light on the microarchitectural

effects of anti-resorption or anabolic drugs used to treat osteoporosis

[10–13].

Characterization of trabecular bone microarchitecture requires a

high resolution, of about 10 µm, in order to obtain suf ficient detail in a

large field of view [14,15]. Two methods are currently available for

Bone 43 (2008) 203–208

⁎ Corresponding author. Fax: +33 238 744 024.

E-mail address: [email protected](C. Chappard).

8756-3282/$ – see front matter © 2008 Elsevier Inc. All rights reserved.

doi:10.1016/j.bone.2008.02.019

Contents lists available at ScienceDirect

Bone

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / b o n e

mailto:[email protected]

http://dx.doi.org/10.1016/j.bone.2008.02.019

http://www.sciencedirect.com/science/journal/87563282

http://www.sciencedirect.com/science/journal/87563282

http://dx.doi.org/10.1016/j.bone.2008.02.019

mailto:[email protected]



assessing bone microarchitecture at this resolution, namely conven-

tional 2D histomorphometry and 3D micro-computed tomography

(µCT). Histomorphometry parameters are usually measured on iliac

crest biopsies embedded in methylmethacrylate, cut into sections

about 5 µm in thickness, and examined under the microscope. Theimages thus obtained are secondly numerized. Standardized defini-

tions of histomorphometric parameters for assessing trabecular bone

microarchitecture were developed in 1987 [16]. The ratio of bone

voxels over tissue voxels was used to determine BV/TV. Mean

intercept length (MIL) analysis is used to determine the ratio of

bone surface over bone volume (BS/BV), trabecular thickness (Tb.Th),

trabecular number (Tb.N), and trabecular separation (Tb.Sp).

However, histomorphometric parameters vary markedly within a

given iliac crest specimen [17]. Moreover, high intraobserver varia-

bility has been reported, suggesting that a large number of sections

from each biopsy should be examined [18]. Finally, histomorpho-

metric measurements on 2D slices are used to estimate 3D structures,

under the assumption that the bone trabeculae are either rod-like or

plate-like in shape. New µCT methods, in contrast, directly provide 3Dinformation without requiring assumptions about the shape of the

trabeculae [19]. Theoretical considerations indicate that stereological

parameters (e.g., connectivity) and degree of anisotropy cannot be

accurately assessed from 2D data [20]. With conventional µCT, micro-

finite element analysis of 3D images can be used to assess the

mechanical properties of cancellous bone from the iliac crest [21,22].

3D microarchitectural analysis of µCT images has been used to assess

the effects of bisphosphonates [12,23] or teriparatide on human bone

[13,24]. A major drawback is that iliac crest biopsy is an invasive

procedure and that paired biopsies must be taken to evaluate

responses to pharmacotherapy, one at baseline from one side and

the other after treatment from the other side. Consequently, most

studies included only small numbers of patients [24]. Furthermore,

the interpretation of paired biopsies requires knowledge of the

heterogeneity in bone microarchitecture within each iliac crest and

between the right and left iliac crests in the same individual.

The objective of this study was to evaluate microarchitectural

heterogeneity within an iliac crest biopsy and between biopsies from

the right and left iliac crests in the same individual. To this end, we

examined iliac crest biopsies from cadavers of postmenopausal

women, comparing within-side and side-to-side values, and taking

measurement reproducibility into account.

Materials and methods

Bone samples

Paired iliaccrest biopsies were harvested in thepost-mortemroom from30 women

with a mean age±standard deviation of 73.7±10.1 years (range, 57–93). The cause of

death, medical history, and prior treatments were unknown.To replicate the conditions

of clinical practice, we used an 8-mm Bordier trephine to collect transilial biopsies. The

trephine was inserted 2 cm under the upper edge of the crest and 2 cm behind the

anterior superior iliac spine [3].

The bone marrow was removed from the biopsies by a chemical defatting

procedure involving submersion in bleach followed by several cycles of submersion in

dichloromethane [25]. Defatted bone contains only bone and air, which facilitates the

critical stepof thresholddetermination.We previously reported thatmicroarchitectural

parameters obtained by conventional X-ray µCT imaging of defatted bone correlated

closely with those obtained by 3D synchrotron radiation µCT [25].

By visual examination, we identified fractured biopsies, biopsies containing

residual bone marrow, and biopsies that did not have two intact cortices. Thesebiopsies were excluded from the study, which left 57 biopsies from 30 individuals,

including 27 biopsy pairs.

Image acquisition

We used the cone-beam Skyscan® 1072 system (Skyscan, Kontich, Belgium) with a

scintillator coupled to a 2D 1024⁎1024 CCD camerawith a dynamic range of 12 bits. To

remove soft X-rays, a 1 mm-thick aluminum filter was placed in front of the X-ray

source, and images were acquired at 80 kV and 100 µA. Reference images without

specimens were taken before each acquisition for flat field correction. Angular rotation

was increased in steps of 0.45°, and at each step two images were acquired (5.8 s per

image) and the two frames were averaged. In all, we acquired 413 radiographic

projections per biopsy, which required about 2 h. 3D images were reconstructed with a

voxel size of 10.77 µm, using the manufacturer's reconstruction algorithm, which is

based on a customized Feldkamp algorithm. This imaging protocol was previously

validated by comparison to synchrotron radiation µCT [25]. Given that voxel resolution

is 10.77 µm with a 1024 ⁎1024 CCD detector matrix, images could not exceed 1 cm in

height; therefore, we performed two acquisitions for large biopsies.

Image processing

For the biopsies that required two acquisitions (n =22/57, 38%), reconstruction was

achieved by visually identifying the section that was identical in the two acquisitions.

Bimodal histograms were obtained in all sections with one peak corresponding to

bone and one to background. The 3D reconstructed volumes were segmented using a

global threshold above which all voxels were classified as bone and below which all

voxels were classified as background. The minimum values between the two peaks

Fig. 1. Examples of histograms of micro-computed tomography reconstructions. The

minimum gray level between the bone and background peaks ranged from 190 to 200.

To separate bone from background in a given biopsy, we used 190 or 200, whichever

was closest to the minimum value in that biopsy.

Fig. 2. 3D reconstruction of a biopsy with cortical bone in dark grey.

204 C. Chappard et al. / Bone 43 (2008) 203– 208



(bone and background) ranged from 190 to 200 (25% of specimens had values close to

200).

For each biopsy, we used a threshold of 190 or 200, whichever was closer to the

minimum value (Fig.1). The same threshold was used for the two biopsies in each pair.

The volume of interest (VOI) was a cylinder 7 mm in diameter high as 75% of the

smallest of the two biopsies. We did not use a filter to reduce image acquisition noise.

3D reconstructions of representative samples are shown in Fig. 2.

Quanti fication of 3D microarchitecture

Trabecular bone

Morphological parameters were calculated on 3D binarized volumes, using

software developed by Skyscan®. Bone surface area (BS) and Bone volume (BV) were

calculatedusing themarching cubes method [26]. Trabecular number (Tb.N, mm−1) was

computed as theinverse of the mean distance between the middle axes of the structure

[27]. Trabecular thickness (Tb.Th⁎, µm) and trabecular spacing (Tb.Sp⁎, µm) were

determined using the model-independent method described by Hildebrand and

Rüegsegger [28].

The degree of anisotropy (DA, no unit), which defines the orientation of the

structure, was derived from the mean intercept length method [29].

The structure model index (SMI) was previously described by Hildebrand [30] and

the trabecular pattern factor (Tb.Pf) was determined as described by Hahn [31]. Both of them were calculated from BV and BS before and after 1-pixel dilation of the 3D model.

SMI quantifies the relative amounts of plates and rods and Tb.Pf is an index of

connectivitybased on the relativeratio betweenconcavity and convexity of the surface.

Cortical bone

Cortical bone was separated from trabecular bone, using visual criteria [32]. The

operator manually adjusted the cortical VOI by meticulously following the edges of the

cortices. Cortical thickness was computed from the VOI and the diameter (d) measured

on a projection free of superimpositions. Each cortex was considered to be a cylinder,

according to the following formula:

Corticalthickness ¼ Volume of the cortex=P4 d=2ð Þ2

ð1Þ

Morphological parameters were derived using Skyscan CTan® software. Cortical

porosity (pore volume over total cortical volume, PoV/TV) was computed. Pore

diameter (PoDm, µm) distribution was computed using the sphere method developed

by Hildebrand and Ruegsegger, as described by Basillais et al. [33].

Side-to-side variability

To evaluate variability in trabecular bone between the right and left iliac crests in

the same individual, we selected the same VOI on both sides, corresponding to about

75% of the height of the smaller iliac crest. We used the same method to compare the

cortices. Data from the two sides were checked for normal distribution then compared

using paired t -tests. The difference was computed as a percentage according to the

following equation:

hΔboneparameter i ¼100

n

Xi¼n

i¼1

jboneparameter rightið Þ À boneparameter i leftið Þj

boneparameter rightið Þð2Þ

Within-side variability

To evaluate variability within each specimen, the whole volume of trabecular bone

from one cortex to the other was divided into three parts containing the same number

of slices. Theboundary betweentrabecular and corticalbone wasidentified visuallyand

was then meticulously followed to draw the VOIs. The two peripheral parts adjacent to

the cortices, were compared to the middle part and each other using Eq. (2) to

determinethe meandifferencein absoluteindividual percentvariation (mAbsΔind). The

three parts were compared using ANOVA when the data were normally distributed.

When ANOVA showed a significant difference, a post-hoc Newman Keuls multiple-

comparison test was performed. For nonnormally distributed data, the Kruskal–Wallis

test was used, with a post-hoc Scheffe's test for one-to-one comparison. All statistical

tests were performed using NCSS® software (.NCSS, Kaysville, UT).

Reproducibility of measurements

Reproducibility of the entire measurement sequence was assessed by performing

two acquisitions for each of 16 biopsies. Visual site matching was used to select the

same VOI for the two acquisitions. We determined BV/TV, BS/BV, Tb.N, Tb.Th⁎, Tb.Sp⁎,

Tb.Pf, SMI, and DA for each acquisition. The root mean square coef ficient of variation

(RMSCV, %) was computed as follows [34]:

RMSCV k ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiPmi¼1

di2=2m

s P x1 þ

P x2=2

Â 100 ð3Þ

where di was the difference between the first and second measurements; m the

number of specimens; and x ̄1 and x ̄2 the means of all first and second measurements,respectively.

Results

Side-to-side variability

Side-to-side variability was assessed using the 27 pairs of biopsies.

Results are reported inTable 1 fortrabecular boneand Table 2 for cortical

bone. Only SMI and cortical thickness had nonnormal distributions. No

statistically significant differences were found between therightand left

side for any of the trabecular or cortical parameters. All correlation

coef ficients between the two sides were significant; r values ranged

from 0.39 to 0.84 for trabecular bone parameters and from 0.43 to 0.61

for cortical bone parameters. For trabecular bone, percent differences

Table 1

Side-to-side variability of trabecular bone parameters in the iliac crest

n = 27 BV/TV a BS/BV b Tb.Th⁎c Tb.Sp⁎d Tb.Ne Tb.Pf f SMIg DAh

Right 14.3 ±4.4 23.8± 4.2 161 ± 30 822 ± 127 0.88± 0.17 1.68± 2.67 1.01± 0.34 1.61± 0.24

Left 14.7 ±4.3 23.7± 4.9 164 ± 35 795 ± 112 0.89± 0.21 2.08± 2.41 1.06 ±0.28 1.60± 0.28

r 0.84** 0.71** 0.64** 0.45* 0.54* 0.73** 0.39*♦ 0.56*

mAbsΔind (%) 14.2 10.8 13.0 11.1 14.8 74.1 84.4 10.5

Data are means±SD on each side, Pearson correlation coef ficients (r ), and the percent differences between the two sides.

**P b

0.001.*P b0.05.♦Spearman correlation coef ficient.

a bone volume/tissue volume, %.b bone surface/bone volume, mm−1.c trabecular thickness, µm.d trabecular spacing, µm.e trabecular number, mm−1.f trabecular pattern factor.g structure model index.h degree of anisotropy.

Table 2

Side-to-side variability of cortical bone parameters in the iliac crest

n = 25 Cortical thickness, μ m Cortical porosity, % Pore diameter, µm

Right 474± 179 42.1 ±9.7 345 ± 49

Left 525 ± 253 40.8± 9.6 350 ± 54

r 0.61♦⁎⁎ 0.54* 0.43*

mAbsΔind (%) 12.3 15.1 11.6

Data are means±SD on each side, Pearson correlation coef ficients (r ), and the percent

differences between the two sides.

**P b0.001.

*P b0.05.♦

Spearman correlation coef ficient.

205C. Chappard et al. / Bone 43 (2008) 203– 208



ranged from 10.5% to 14.8% for all parameters except Tb.Pf (74.1%) and

SMI (84.4%). For cortical bone, differences were 2.3% for pore diameter

and 12.3% for cortical thickness.

Within-side variability

Within-side variability was evaluated using all 57 biopsies. Results

are reported in Table 3. Distribution was normal in each of the three

parts for three parameters: BV/TV, BS/BV, and Tb.Th⁎. Statistically

significant differences between the three parts were noted only for

BV/TV and DA; post-hoc testing established that the difference was

between the middle and the peripheral parts for both parameters. A

difference of borderline significance was found for Tb.Sp⁎ (P =0.06).Except for Tb.Pf, mean individual percent differences relative to the

middle part ranged from 8.4% (Tb.Sp⁎) to 29.3% (SMI). For Tb.Pf,

extremely high percent differences of up to 197% were found.

Reproducibility of measurements

RMSCV(%) was 5.5% for BV/TV, 4.3% for BS/BV, 2.6% for Tb.Th⁎, 6.3%

for Tb.N, 4.1% for Tb.Sp⁎, 22.1% for Tb.Pf, 6.9% for SMI, and 4.5% for DA.

Discussion

When studying bone biopsies, side-to-side variability, within-side

variability, and reproducibility of the measurement sequence are

important to consider, most notably in longitudinal studies. This workis the first to address all three of these points. We found that Tb.Sp⁎ and

Tb.Th⁎were robust parameters characterized by limited side-to-side and

within-side variability. Consequently, they hold the greatest promise for

longitudinal studies. To a lesser extent, BV/TV, BS/BV, Tb.N, and DA may

be useful. In contrast, SMI and Tb.Pf showedconsiderable variability that

may limit their relevance.

The side-to-side correlation was strongest for BV/TV (%) (r =0.84).

Intermediate correlations were noted for BS/BV (r = 0.71), Tb.Th⁎

(r =0.64), and Tb.Pf (r =0.73). Except for Tb.Pf and SMI, the mean

individual percent difference (mAbsΔind) was acceptable, the range

being 10.8% to 14.2%. For the usual morphological parameters, RMSCV%

estimated from two different acquisitions of the same biopsy ranged

from 4.1% to 6.3%. In an earlier study, differences in VOI positioning of up

to 10 slices in either direction resulted in less than 1.4% variation [25].

Reproducibility of our 3D analysis wasabout 2-fold thatof conventional

histomorphometry: this, coef ficients of variation of usual microarchi-

tectural measurements ranged in an earlier study from 2% to 6% in the

intraobserver analysis and from 1.8% to 7.1% in the section-to-section

analysis [18]. The good reproducibility of our measurements indicates

that any differences found between the right and left sides reflect true

differences, as opposed to measurement error related to the acquisition

and analysis sequence. Ideal parameters for longitudinal studies must

exhibit both limited side-to-sidevariability and good reproducibility. Tb.

Th⁎ seems to be a robust parameter, with only about 2.6% of variability

between two acquisitions. Reproducibilitycan be considered acceptable

for Tb.Sp⁎, DA, and BS/BV; moderate for BV/TV, Tb.N, and SMI; and poor

for Tb.Pf. Tb.Th⁎ is an interesting parameter with side-to-side andwithin-side variabilities of about 13%, i.e., about 5 times the variability

found in the reproducibility assessment. In contrast, TbPf and SMI

showed considerable side-to-side variability and poor reproducibility.

However, consideringthe satisfactory correlationbetween the right and

left sides, Tb.Pf may hold promise, although improvements in the

algorithm are probably needed.

BV/TVand DAwerelowerin the middleof the iliac crest thanin the

juxtacortical regions, the difference being as large as 23.5% for BV/TV.

When comparing the middle to the juxtacortical regions, trends were

found toward lower values for Tb.Th⁎ and Tb.N and higher values for

BS/BV and Tb.Sp⁎. The differences between juxtacortic.a and juxta-

cortic.b were of the same order than between juxtacortic.a/b and mid.

due to a larger variability of trabecular bone parameters close to

cortices. These findings indicate a need for standardizing the distancefrom the cortex when selecting subsamples of iliac crest biopsies. The

variation of bone histomorphometric trabecular bone parameters

have already been assessed in two contiguous biopsies and depended

on diagnosis and parameters; for example in osteoporosis difference

could reach 30% for trabecular bone volume [35].

We designed an experimental protocol that ensured optimal

conditions for µCT analysis. We chose a high voxel resolution of

10.78 μ m based on the good compromise thus obtained between field

of view (1024⁎1024pixels) andbone samplesize, thegood agreement

between our desktop µCTand synchrotron radiation µCTanalyses [25],

and recommendations regarding prediction accuracy [14,15]. We used

defatted human bone, whose quasi-biphasic structure composed of

trabecular bone and background with no marrow simplifies threshold

determination. The values of structural indices are dependent on the

Table 3

Within-side variability of trabecular bone parameters across three parts of identical height, namely, the two juxtacortical parts and the middle part

n =57 TV a BV/TV b BS/BV c Tb.Th⁎d Tb.Sp⁎e Tb.Nf Tb.Pf g SMIh DAi

1/3 juxtacort.a 67.8 ±18.7 17.6 ±5.6 22.6 ±3.2 173.0 ±40 692 ±96 1.01 ±0.2 2.40 ±2.54 1.01 ±0.29 1.82 ±0.34

(22.8–108.3) (6.1– 31.8) (14.6 – 35 .7 ) ( 10 0– 266 ) ( 49 3–925) (0.61–1.48) (− 3.65–8.96) (0.47 –1.69) (1.25– 2.78)

1/3 middle 79.6 ±21.5 14.9 ±4.2 24.2± 4.5 157.0 ±31 711± 110 0.95 ±0.2 3.03± 2.5 1.15± 0.46 1.64 ±0.32

(21.3–136) (5.7 – 26.3) (16.7 – 39 .1) ( 88– 229) (511–1026) (0.56 –1.55) (− 3.04–10.94) (0.50– 3.80) (0.69– 2.69)

1/3 juxtacort.b 71.8. ±20.5 17.2 ±5.6 23.0 ±5.1 170 ±38.7 699 ±109 1.00 ±0.2 2.26 ±3.1 1.10 ±0.34 1.80 ±0.35

(24.5–116.2) (5.5– 29.7) (14.2– 38 .8 ) ( 95– 266) (425–991) (0.58–1.58) (−5.6 –10.5) (0.37 –1.84) (1.14– 2.70)

ANOVA 0.006# 0.01# ns ns ns♦ ns♦ ns♦ ns♦ 0.004♦#mAbsΔind (%) juxtacort.a/mid. 16.7 22.9 13. 3 16.2 7.6 12.5 107 16.8 15.3

mAbsΔind (%) juxtacort.b/mid. 13.5 23.5 12.3 15.8 8.1 10.9 200 22.0 15.9

mAbsΔind (%) juxtacort.a/ juxtacort.b 18.3 26.2 10.8 20.8 8.8 11.2 118 21.2 16.7

Data are means±SD with the range in italics.♦Kruskal–Wallis test instead of ANOVA.

#With a post-Hoc test juxtacort.a/mid. and juxtacort;b/mid. are significantly different.a Total volume.b bone volume/tissue volume, %.c bone surface/bone volume, mm−1.d trabecular thickness, µm.e trabecular spacing, µm.f trabecular number, mm−1.g trabecular pattern factor.h structure model index.i degree of anisotropy.




threshold used. In a previous study, we used a single threshold

corresponding to the mean minimum value between peaks of bone

and background for the whole data set from femoral head samples

[25]. Here, we used two threshold values, 190 and 200, given the

considerable variations in trabecular bone appearance and histo-

grams. The mean minimum value between peaks ranged from 190 to

200, and for each sample we therefore chose the threshold value,190

or 200, closest to the mean minimum value in that sample. Never-

theless, we used the same threshold for the right and left sides of agiven individual, in order to standardize the side-to-side comparison.

The iliac crest samples had a lower signal-to-noise ratio with a mean

BV/TV of 14.7% compared to 22.1% for the femoral head samples

examined in our previous study under similar acquisition conditions

[25]. Using a single threshold such as 190 for the entire data set would

probably have resulted in underestimation of all the trabecular bone

parameters from specimens with a minimum value close to 200.

In a longitudinal study comparingiliac crest biopsies beforeand after

3 years of risedronate or placebo in postmenopausal women, marked

bone loss occurred in the placebo group, with a 28% decrease in BV/TV%

and a 20% decrease in Tb.Th⁎. In the higher turnover risedronate group,

changes in BV/TV% ranged from 21.9% to 22.7%, with a mean annual

change of 5.6% [12]. Inour study, meanBV/TV was 14.3%on the right side

and 14.7% on the left side; the side-to-side difference of 14.2% was the

mean of the absolute individual variations, as opposed to the percent

difference in mean values between the right and left sides. This percent

difference was 0.4%, i.e., more than 20 times less than the difference

under treatment described in the above mentioned study.

The mean age of our population was 73.7±10.1 years (range, 55–

96 years), which was older than the mean age in most of the clinical

studies. In the study comparing risedronate to a placebo [12], mean age

was 65.5 years in both groups. Mean age was 54 years in a study of PTH

therapy in 8 women [24] and 66 years in a study of alendronate versus

placebo in osteoporotic women [23]. In a study comparing iliac bone

before and after the menopause, mean age was 49 years at the first

biopsy and 54.6 years at the second biopsy [36]. In our cadaver study,

mean agewasolder, in keepingwiththe mean life expectancyof 83 years

for women in France. According to French law, we had no information

about diseases or medication use in our study individuals. We excludedbiopsies that were visuallyabnormal, includingone biopsythat probably

contained a bone metastasis and another with considerable hetero-

geneity consistent with the presence of a microcallus. Although our

populationcannot be considered normal regarding bone health,it wasa

typical cross-section of women who died at a mean age of 74 years for

unknown reasons. Given the invasivenessof iliac crest biopsy, studies of

normal individuals must often rely on cadaver biopsies [3].

Microarchitectural parameter values varied considerably across

studies. In postmenopausal women, BV/TV values ranged from 13% to

31% [10–14,18,23,24,27,36]. For instance, mean BV/TV was 21.3% in 8

women shortly after the menopause [24] and 25.1% in another study of

early postmenopausal women [11]. Most studies designed to assess

treatment effects used resolutions of about 30 µm [11–13,24], that is,

about three times higher than in our study, which might contribute toexplain the differences with our results [37]. In a population similar to

ours, with the same methods of data acquisition and analysis, results

were consistent with our findings [27]: in 62 iliac crest biopsies from

men and women having a mean age of 69± 15 years, with 14 µm

resolution, mean values were 15.2% for BV/TV,150 µm for Tb.Th, 754 µm

for Tb.Sp,1.39mm−1 forTb.N, and1.5 forDA. Except fora small difference

in Tb.N, these values closely match those found in our study. The only

noticeable difference with our methodology was that fresh biopsies

placed in sample holders with formaldehyde were used, compared to

defatted bone in our study. The consistency of results between two

studies characterizedby similar populations, acquisitionparameters, and

analysis methods suggests that these results may have general validity.

Several methods have been used to compute cortical thickness,

including determination of the mean thickness of the 3D cortex using

the sphere method [24] and use of a Euclidian distance map [11]. Mean

cortical thickness was 632 µm in early postmenopausal women [24],

compared to about 500 µm in our study. The sphere method is suitable

only for binarized images, and when using the Euclidian distance map

method all holes in the cortical region are automatically filled in. With

our method, the measurements were completely independent from the

threshold value. The cortical VOI was defined manually using an

algorithm that allowed the operator to meticulously follow the edges of

the cortices. The main limitation of this method is that it requires intactcortices, since it assumes a cylindrical shape of the cortical region.

The using of defatted bone necessitated to separate biopsies in two

parts: one forhistomorphometryand one for micro-CT analysis. In our

study, it was necessary to use the whole specimen for micro-CT

analysis. However, it would be interesting to compare histomorpho-

metric measurements with micro-CT measurements.

In sum, bone microarchitecture can be reliably evaluated by µCT.

Similar findings can be expected when the populations, acquisition

protocols, and analysis methods are comparable. Within-side varia-

bility is similar in magnitude to side-to-side variability. The robustness

of trabecular bone parameters varies widely, suggesting that studies

may need to focus on the small number of highly robust parameters.

3D µCT analysis can probably supply reliable data in longitudinal

studies based on small number of patients, provided the µCT

acquisition and data analysis methods comply with stringent quality

criteria.

Acknowledgments

I would thank Maurice Arras from the Institut of Anatomie Paris 5

for helping me to harvest iliac crest biopsies and I thank ANR

MIPOMODIM no. R06009NG for financial support.

References

[1] Osteoporosis prevention, diagnosis, and therapy. NIH Consens Statement 2000,vol.17; 2000. p. 1–45. (March 27–29).

[2] Parfitt AM, Mathews CH, Villanueva AR, Kleerekoper M, Frame B, Rao DS.Relationships between surface, volume, and thickness of iliac trabecular bone in

aging and in osteoporosis. Implications for the microanatomic and cellularmechanisms of bone loss. J Clin Invest 1983;72:1396–409.

[3] Aaron JE, Makins NB, Sagreiya K. The microanatomy of trabecular bone loss innormal aging men and women. Clin Orthop Relat Res 1987;215:260–71.

[4] Podenphant J, Gotfredsen A, Nilas L, Norgaard H, Braendstrup O. Iliac crest biopsy:representativity for the amount of mineralized bone. Bone 1986;76:427–30.

[5] Mosekilde L, Viidik A, Mosekilde L. Correlation between the compressive strengthof iliac and vertebral trabecular bone in normal individuals. Bone 1985;65:291–5.

[6] Oleksik A, Ott SM, Vedi S, Bravenboer N, Compston J, Lips P. Bone structure inpatients with low bone mineral density with or without vertebral fractures. J BoneMiner Res 2000;15:1368–75.

[7] Dalle Carbonare L, Arlot ME, Chavassieux PM, Roux JP, Portero NR, Meunier PJ.Comparison of trabecular bone microarchitecture and remodeling in glucocorticoid-induced and postmenopausal osteoporosis. J Bone Miner Res 2001;16:97–103.

[8] Legrand E, Chappard D, Pascaretti C, Duquenne M, Krebs S, Rohmer V, et al.Trabecular bone microarchitecture, bone mineral density, and vertebral fracturesin male osteoporosis. J Bone Miner Res 2000;15:13–9.

[9] Parisien M, Cosman F, Mellish RW, Schnitzer M, Nieves J, Silverberg SJ, et al. Bonestructure in postmenopausal hyperparathyroid, osteoporotic, and normal women.

J Bone Miner Res 1995;10:1393–9.[10] Eriksen EF, Melsen F, Sod E, Barton I, Chines A. Effects of long-term risedronate on

bone quality and bone turnover in women with postmenopausal osteoporosis.Bone 20 02;31:620–5.

[11] Dufresne TE, Chmielewski PA, Manhart MD, Johnson TD, Borah B. Risedronatepreserves bone architecture in early postmenopausal women in 1 year as measuredbythree-dimensional microcomputedtomography. CalcifTissue Int2003;73:423–32.

[12] Borah B, DufresneTE, ChmielewskiPA, JohnsonTD, Chines A, ManhartMD. Risedronatepreserves bone architecture in postmenopausal women with osteoporosis asmeasured by three-dimensional microcomputed tomography.Bone 2004;34:736–46.

[13] Jiang Y, Zhao JJ, Mitlak BH, Wang O, Genant HK, Eriksen EF. Recombinant humanparathyroid hormone (1–34) [teriparatide] improves both cortical and cancellousbone structure. J Bone Miner Res 2003;18:1932–41.

[14] Muller R, Koller B, Hildebrand T, Laib A, Gionollini S, Rüegsegger P. Resolutiondependency of microstructural properties of cancellous bone based on threedimensional μ -tomography. Technol Health Care 1996;4:113–9.

[15] Peyrin F, Salome M, Cloetens P, Laval-Jeantet AM, Ritman E, Ruegsegger P. MicroCTexaminations of trabecularbonesamples at differentresolutions: 14, 7 and2 micron

level. Technol Health Care 1998;6:391–

401.

207C. Chappard et al. / Bone 43 (2008) 203– 208



[16] Parfitt AM, Drezner MK, Glorieux FH, Kanis JA, Malluche H, Meunier PJ, et al. Bonehistomorphometry: standardization of nomenclature, symbols, and units. Reportof the ASBMR Histomorphometry Nomenclature Committee. J Bone Miner Res1987;2:595–610.

[17] Parisien MV, McMahon D, Pushparaj N, Dempster DW. Trabecular architecture iniliac crest bone biopsies: intra-individual variability in structural parameters andchanges with age. Bone 1988;9:289–95.

[18] Podenphant J, Gotfredsen A, NilasL, NorgardH, Braendstrup O, Christiansen C. Iliaccrest biopsy: an investigation on certain aspects of precision and accuracy. BoneMiner 1986;1:279–87.

[19] Muller R. Bone microarchitecture assessment: current and future trends.

Osteoporos Int 2003;14(Suppl 5):89–

99.[20] Serra J. Image analysis and mathematical morphology. London: Academic Press;1982. 104 pp.

[21] vanRietbergenB, Majumdar S, Pistoia W, Newitt DC, KothariM, LaibA, RuegseggerP. Assessment of cancellous bone mechanical properties from micro-FE modelsbased on micro-CT, pQCT and MR images. Technol Health Care 1998;6:413–20.

[22] Uchiyama T, Tanizawa T, Muramatsu H, Endo N, Takahashi HE, Hara T. Three-dimensional microstructural analysis of human trabecular bone in relation to itsmechanical properties. Bone 1999;25:487–91.

[23] Recker R, Masarachia P, Santora A, Howard T, Chavassieux P, Arlot M, et-al. Trabecularbonemicroarchitecture after alendronatetreatmentof osteoporotic women.Curr MedRes Opin 21: 185–194. (Erratum in: Curr Med Res Opin 2005;21:324).

[24] Dempster DW, Cosman F, Kurland ES, Zhou H, Nieves J, Woelfert L, et al. Effects of daily treatment with parathyroid hormone on bone microarchitecture andturnover in patients with osteoporosis: a paired biopsy study. J Bone Miner Res2001;16:1846–53.

[25] ChappardC, BasillaisA, Benhamou L, BonassieA, Brunet-ImbaultB, Bonnet N,et al.Comparison of synchrotron radiation and conventional X-ray microcomputedtomography for assessing trabecular bone microarchitecture of human femoral

heads. Med Phys 2006;33:3568–77.

[26] Lorensen WE, Cline HE. Marching cubes: a high resolution 3D surface constructionalgorithm. Comput Graph 1987;21(Suppl 1):7–12.

[27] Ulrich D, Van Rietbergen B, Laib A, Ruegsegger P. The ability of three dimensionalstructural indicesto reflect mechanicalaspectsof trabecular bone.Bone 1999;25:55–60.

[28] Hildebrand T,RuegseggerP. A newmethodfor themodel independent assessmentof thickness in three dimensional images. J Microsc 1997;185:67–75.

[29] Whitehouse WJ. The quantitative morphology of anisotropic trabecular bone. J Microsc 1974;101:153–6.

[30] Hildebrand T, Ruegsegger P. Quantification of bone microarchitecture with thestructure model index. Comput Methods Biomech Biomed Engin 1997;1:15–23.

[31] Hahn M, Vogel M, Pompesius-Kempa M, Delling G. Trabecular bone pattern

factor—

a new parameter for simple quantification of bone microarchitecture.Bone 1992;13:327–30.[32] Dufresne T. Segmentation techniques for analysis of bone by three-dimensional

computed tomographic imaging. Technol Health Care 1998;6:351–9 (Review).[33] Basillais A, Bensamoun S, Chappard C, Brunet-Imbault B, Lemineur G, Ilharreborde

B, et al. Three dimensional characterization of cortical bone microstructure bymicro-computed tomography: validation with ultrasound and microscopicmeasurements. J Orthop Sci 2007;12:141–8.

[34] Chappard C, Kolta S, Fechtenbaum J, Dougados M, Roux C. Clinical evaluation of spine morphometric X-ray absorptiometry. Br J Rheum 1998;37:49–51.

[35] Chavassieux PM, Arlot ME, Meunier PJ. Intersample variation in bone histomor-phometry: comparison between parameter values measured on two contiguoustransiliac bone biopsies. Calcif Tissue Int 1985;37:345–50.

[36] Akhter MP, Lappe JM, Davies KM, Recker RR. Transmenopausal changes inthe trabecular bone structure. Bone 2007;41:111–6 (Electronic publication 2007Apr 10).

[37] Kim DG, Christopherson GT, Dong XN, Fyhrie DP, Yeni YN. The effect of microcomputed tomography scanning and reconstruction voxel size on theaccuracyof stereological measurementsin human cancellous bone. Bone2004;35:

1375–82.


3d bone micro archit

Documents