- Email: [email protected]
Age-related changes of vertical and horizontal lumbar vertebral trabecular 3D bone microstructure is different in women and men
Bone 57 (2013) 47–55
Contents lists available at ScienceDirect
Bone journal homepage: www.elsevier.com/locate/bone
Technical Note
Age-related changes of vertical and horizontal lumbar vertebral trabecular 3D bone microstructure is different in women and men Jesper Skovhus Thomsen ⁎, Andreas Steenholt Niklassen, Ebbe Nils Ebbesen, Annemarie Brüel Department of Biomedicine — Anatomy, Aarhus University, Aarhus, Denmark
a r t i c l e
i n f o
Article history: Received 9 April 2013 Revised 15 July 2013 Accepted 16 July 2013 Available online 27 July 2013 Edited by: David Burr Keywords: Bone μCT Bone structure Static histomorphometry Human vertebra Sex differences Aging
a b s t r a c t The study presents a 3D method for subdividing a trabecular network into horizontal and vertical oriented bone. This method was used to investigate the age related changes of the bone volume fraction and thickness of horizontal and vertical trabeculae in human lumbar vertebral bone estimated with unbiased 3D methods in women and men over a large age-range. The study comprised second lumbar vertebral body bone samples from 40 women (aged 21.7–96.4 years, median 56.6 years) and 39 men (aged 22.6–94.6 years, median 55.6 years). The bone samples were μCT scanned and the 3D microstructure was quantified. A voxel based algorithm inspecting the local neighborhood is presented and used to segment the trabecular network into horizontal and vertical oriented bone. For both women and men BV/TV decreased significantly with age, Tb.Th* was independent of age, while SMI increased significantly with age. Vertical (BV.vert/TV) and horizontal (BV.horz/TV) bone volume fraction decreased significantly with age for both sexes. BV.vert/TV decreased significantly faster with age for women than for men. Vertical (Tb.Th*.vert) and horizontal (Tb.Th*.horz) trabecular thickness were independent of age, while Tb.Th*.horz/Tb.Th*.vert decreased significantly with age for both sexes. Additionally, the 95th percentile of the trabecular thickness distribution increased significantly with age for vertical trabeculae in women, whereas it was independent of age in men. In conclusion, we have shown that vertical and horizontal oriented bone density decreases with age in both women and men, and that vertical oriented bone is lost more quickly in women than in men. Furthermore, vertical and horizontal trabecular thickness were independent of age, whereas the horizontal to vertical trabecular thickness ratio decreased significantly with age indicating a relatively more pronounced thinning of horizontal trabeculae. Finally, the age-related loss of trabecular elements appeared to result in a compensatory hypertrophy of vertical trabeculae in women, but not in men. © 2013 Elsevier Inc. All rights reserved.
Introduction During aging, human vertebral bone is lost resulting in weaker bone and thereby a higher fracture risk. It has been shown that bone density is a major determinant of vertebral bone strength [1,2], but it has also been suggested that the microstructure of the bone tissue plays a role for the bone strength [3–6]. Hence, Hui et al. showed that even for constant bone mass, fracture risk increases with age [7]. Thus, bone fracture strength is not only dependent on bone density, but also on bone microstructure, micro damage accumulation, and mineralization [8]. The conventional view is that a compressive load on a vertebral body is mainly carried by the vertical trabeculae, whereas the horizontal trabeculae serve to prevent buckling of the vertical trabeculae [9–12]. This view is reinforced by finite element analyses of human vertebral bone ⁎ Corresponding author at: Department of Biomedicine — Anatomy, Aarhus University, Wilhelm Meyers Allé 3, DK-8000 Aarhus C, Denmark. E-mail addresses: [email protected] (J.S. Thomsen), [email protected] (A.S. Niklassen), [email protected] (E.N. Ebbesen), [email protected] (A. Brüel). 8756-3282/$ – see front matter © 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.bone.2013.07.025
specimens, which demonstrates that vertical trabeculae are more highly strained than horizontal trabeculae under normal compressive loading [13–16]. Furthermore, Fields et al. hypothesized that vertebral bone strength is better explained by the bone volume fraction of the vertical trabeculae alone, than by the bone volume fraction of all trabeculae [15]. In a subsequent study, Fields et al. found that the vertical trabeculae played a particular important role for the compressive bone strength of vertebrae with low bone density [17]. Consequently, it is important to quantify the age-related changes in trabecular thickness as well as bone volume fraction for horizontal and vertical trabeculae separately. It has, for a long time, been debated whether the removal of vertical trabeculae with age would lead to a higher compressive load on the remaining trabeculae, and thereby to a compensatory thickening of these [12,18–22]. Thus, in 1983 Parfitt et al. suggested that, when trabeculae are removed during aging, the remaining trabeculae would be more widely separated and therefore might undergo compensatory thickening [23]. Furthermore, in a 1999 perspective article Harold Frost invited other researchers to provide proof that vertical trabeculae can strengthen and ‘thicken’ in adults under increased mechanical
48
J.S. Thomsen et al. / Bone 57 (2013) 47–55
loading [18]. We have previously investigated this using 2D histomorphometry on human vertebral bone, but were unable to find evidence of a compensatory thickening of vertebral trabeculae with age [24]. However, the 2D analysis was limited by the use of the parallel plate model for determination of the trabecular thickness. The development of micro computed tomography (μCT) scanners has now made it feasible to investigate the microstructure of the 3D trabecular network in human vertebrae using model assumption free methods for determination of the trabecular thickness. Therefore, the aim of the present study was to present a 3D method for segmenting the trabecular network into horizontal and vertical oriented bone. Furthermore, we applied the method to 79 human vertebral bone specimens and tested the hypothesis that the age related changes of the bone volume fraction and thickness of horizontal and vertical trabeculae in human vertebral bone do not differ in women and men. Finally, we investigated whether vertical vertebral trabeculae can thicken with age. Materials and methods Bone specimens Second lumbar vertebral bodies from 40 women (aged 21.7– 96.4 years, median 56.6 years) and 39 men (aged 22.6–94.6 years, median 55.6 years) were analyzed. The subjects were chosen so as to provide an even age and sex distribution in the 20–90 years age range. All individuals were Caucasian. Vertebral bodies with fractures identified from lateral-projection X-ray images were excluded from further investigations as previously reported [1]. Samples from individuals with known cancer or drug abuse were excluded from the study at the time of autopsy. Subsequently, a thorough review of hospital records and autopsy information available was used to exclude individuals with cancer (discovered at autopsy), metabolic disease, severe liver or kidney disease, medication affecting bone metabolism, or periods of more than 2 weeks of immobilization prior to death as previously described [25]. The vertebral bodies from 24 of the women and 24 of the men have previously been investigated using 2D histomorphometry [24,26]. The investigation presented here is a subset of a larger study called the “Danish in Vitro Bone Study” (DAVIBO) [1,25,27]. The collection of the material and the study design were approved by the local ethical committee. Specimen preparation
was interactively delineated as closely as possible to either the cortical bone or the sawing planes using custom-made software written by one of the authors (JST) [29]. The resulting image masks were transferred back to the μCT scanner and imported into IPL (version 5.11, Scanco Medical). The 3D data sets were low-pass filtered using a Gaussian filter (σ = 0.8, support = 1) in order to remove noise and were subsequently segmented with a fixed threshold filter. The minimum point between the marrow and the bone peak in the attenuation histogram was automatically determined using IPL for 25 vertebral bone specimens, and the median of these thresholds (450.7 mg HA/cm3) was used for all segmentations. Standard quantification of the microstructure of the trabecular bone network was performed using the software provided with the μCT scanner (version 6.0, Scanco Medical). Microstructural measures included bone volume per total volume (BV/TV), trabecular thickness (parallel plate model: Tb.Th [30] and directly estimated: Tb.Th* [31]), trabecular number (Tb.N*) [32], connectivity density (CD) [33], and structural model index (SMI) [34]. The computation of these structural measures has previously been described in detail [35]. In addition, the degree of anisotropy (DA) was also computed [36]. Quality assurance was performed by weekly (density) and monthly (geometry) scans of the solid-state calibration phantom provided with the scanner. Trabecular orientation The voxels of the trabecular bone network was classified as either horizontal or vertical by inspecting the orientation of the bone tissue surrounding the voxel. The method is an extension of our previously presented 2D method into 3D [24]. Consider a voxel characterized by its Cartesian coordinates (x0, y0, z0). Let l*(θ, φ) denote the Euclidian length from the voxel at (x0, y0, z0) to either the bone-marrow boundary or the VOI boundary, where θ is the polar angle and φ is the azimuthal angle (Fig. 1). Let
hðx0 ; y0 ; z0 Þ ¼
N −1 X
3N 8 −1
X
j¼0 i¼N8
vðx0 ; y0 ; z0 Þ ¼
N N −1 X 8 −1 X
j¼0 i¼0
i j l 2π ; 2π N N
ð1Þ
NX −1 N −1 X i j i j þ l 2π ; 2π l 2π ; 2π N N N N 3N j¼0
ð2Þ
i¼ 8
The bone specimens had previously been prepared for histomorphometry [28]: The vertebral bodies were cleaned of soft connective tissue and halved along the anterior–posterior axis using a diamond parallel precision saw (Exakt, Apparatebau, Otto Herrmann, Norderstedt, Germany). An approximately 9-mm-thick frontal bone specimen was sawed from the center of one half of the corpus and embedded undecalcified in methyl methacrylate (Technovit 9100, Heraeus Kulzer, Wehrheim/Ts., Germany). Subsequently, the tissue blocks were trimmed using the diamond saw to remove all excessive methyl methacrylate to facilitate μCT scanning. Micro computed tomography (μCT) The bone specimens were placed in a μCT scanner (μCT35, Scanco Medical AG, Brüttisellen, Switzerland) so that they were aligned along the x-axis of the scanner and so the vertebral endplates were parallel with the xy-plane. The specimens were scanned in high-resolution mode (1000 projections per 180°) with an isotropic voxel size of 18.5 μm, an X-ray tube voltage of 70 kVp, an X-ray tube current of 114 μA, and an integration time of 800 ms. The reconstructed μCT data was exported to a PC running Linux (OpenSUSE 11.2, http://www. opensuse.org), where the trabecular bone volume of interest (VOI)
Fig. 1. Coordinate system with the origin placed in the voxel under inspection (x0, y0, z0) used for the calculations. The Euclidian length from the origin to either the bone-marrow boundary or the VOI boundary is denoted l*(θ, φ), where θ is the polar angle and φ is the azimuthal angle.
J.S. Thomsen et al. / Bone 57 (2013) 47–55
where N is a multiple of 8. If h(x0,y0,z0) N v(x0,y0,z0) the voxel at (x0, y0, z0) is classified as horizontal, otherwise it is classified as vertical. In all, l* is determined for N2/2 horizontal and N2/2 vertical directions for each voxel. In the present study N = 128 and consequently l* was determined for 1282 = 16,384 directions for each voxel. Only voxels representing bone inside the VOI is considered, while voxels representing marrow is not considered. In order to avoid single voxel protrusions on the trabeculae as being identified as a separate trabecula, the resulting data set underwent a median filtration. The standard software provided with the Scanco scanner outputs a file containing the “local” Tb.Th* values determined for all bone voxels. The standard software computes the Tb.Th* value for the entire trabecular network as the mean of these local Tb.Th* values. As these thickness values are determined using the direct estimation method they are not biased by model assumptions. Thus, by combining the h and v sets with the Tb.Th* file it is possible to estimate the average thickness of horizontal trabeculae, Tb.Th*.horz, and vertical trabeculae, Tb.Th*.vert, respectively. The h and v data sets were exported and saved for subsequent 3D visualization and further data processing. The above described method was implemented in a custom made computer program written in C and running under Linux using the Tb.Th* file as input. This file serves the multiple purpose of identifying bone voxels inside the VOI (a Tb.Th* value larger than 0) and for computation of Tb.Th*.horz and Tb.Th*.vert. As l*(θ, φ) has to be calculated many times for each voxel, the algorithm for determining l*(θ, φ) has to be very fast. We implemented a 3D version of Bresenham's line algorithm [37] for this purpose, which is normally considered very efficient. From the h and v data sets the horizontal bone volume fraction, BV.horz/TV, and the vertical bone volume fraction, BV.vert/TV, can easily be computed as the number of voxels in either the h or v data set divided by the total number of voxels in the VOI. In addition, the thickness distribution for both the horizontal and the vertical direction was determined and the 50th and the 90th–99th percentiles were determined for these distributions. 3D visualization The h and v data sets were 3D visualized as isosurfaces for selected bone samples using Amira 4.1.2 (Mercury Computer Systems, Inc., Chelmsford, Mass., USA) running under Linux. Statistics Linear regression was performed using the least-squares method using custom made software. Comparisons between two linear relationships y1 = a1 x + b1 and y2 = a2 x + b2 each representing N data points were carried out using a t-test [38]. For comparison between the two slopes, a1 and a2, the test statistic was [39]: ja1 −a2 j ffi t ¼ vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u X 2 X 2 uN y1 − y1 1−r 2 u 1 u t X 2 X 2 N−2 N x − x 1
ð3Þ
Both tests were carried out with ν = N − 2 degrees of freedom. The probability level p for the test statistic t was computed as [40]: t
∫
1þ
−t
pðt Þ ¼ 1−
1
ν
1=2
x2 ν
−1=2
∫x
!−ðνþ1Þ=2 dx ð5Þ ðν=2Þ−1
ð1−xÞ
dx
0
p b 0.05 was considered significant. If the slope of the regression line did not differ between women and men an analysis of covariance (ANCOVA) was performed using the statistical software R (http://www.r-project.org). Results Trabecular orientation Fig. 2A shows an example of the voxel orientation classification algorithm. Horizontal voxels are shown in blue and vertical voxels are shown in red. The efficacy of the algorithm can clearly be seen from the 3D visualization. Figs. 2B and C shows a magnification of the marked region in Fig. 2A. Fig. 2B is computed without median filtration, whereas Fig. 2C is computed with median filtration. It can clearly be seen that the small specs of blue horizontal voxels that can be observed in Fig. 2B, have been filtered out in Fig. 2C without compromising the shape and extent of the trabeculae. Figs. 2D–I show examples of the algorithm for women and men at different ages. Age-related changes in bone architecture Measurements of bone architecture are given in Table 1. BV/TV decreased significantly with age for both women (r = −0.83, p = 3.2 × 10−11) and men (r = −0.74, p = 8.8 × 10−8). These agerelated changes in BV/TV did not differ significantly between women and men. Average trabecular thickness (Tb.Th*) determined using the direct method did not change significantly with age for either women or men. Similarly, average trabecular thickness determined using parallel plate assumptions (Tb.Th) did also not decrease significantly with age for women or men. However, if women and men were pooled Tb.Th decreased significantly with age (r = −0.26, p = 0.019). SMI increased significantly with age for women (r = 0.66, p = 3.5 × 10−6) and men (r = 0.44, p = 0.0054). The fit line was significantly steeper (p = 0.0054) for women than for men, indicating a faster transformation of the trabecular bone network from plate-like to rod-like in women. Tb.N* decreased significantly with age for both women (r = −0.77, p = 5.6 × 10−9) and men (r = −0.63, p = 1.5 × 10−5), and likewise CD decreased significantly with age for both women (r = −0.46, p = 0.0027) and men (r = −0.40, p = 0.011). These age-related changes in Tb.N* and CD did not differ between the sexes. The degree of anisotropy (DA) was independent of age for women (r = 0.094, n.s.), whereas it increased significantly with age for men (r = 0.52, p = 7.7 × 10−4). Age-related changes of horizontal and vertical trabeculae
1
where r1 is Pearson's correlation coefficient. For comparison between the two y-axis intercepts, b1 and b2, the test statistic was [39]: jb1 −b2 j ffi t ¼ vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi uX X 2 X 2 u 2 2 N y − y u x1 1 1 1−r 1 u X 2 X 2 N−2 t N N x1 − x1
49
ð4Þ
The trabecular bone volume fraction of vertical voxels (BV.vert/TV) decreased significantly with age for women (r = −0.82, p = 1.4 × 10−10) and men (r = −0.68, p = 2.1 × 10−6) (Fig. 3A). The slope of the fit line differed significantly (p = 0.014) between women and men, showing that BV.vert/TV decreased significantly faster with age for women than for men. BV.horz/TV also decreased significantly with age for women (r = −0.75, p = 2.5 × 10−8) and men (r = −0.71, p = 3.4 × 10−7), but these age-related difference did not differ between the sexes (Fig. 3B). The fit line for BV.vert/TV was significantly steeper than the fit line for BV.horz/TV for both women
50
J.S. Thomsen et al. / Bone 57 (2013) 47–55
B
A
C
D
G
E
F
H
I
Fig. 2. 3D visualization of the trabecular network subdivided into horizontal (blue) and vertical (red) voxels. (A) 27-year-old man. (B) Magnification of the region shown with a box in (A) without median filtration (see text for detail). Arrowheads indicates regions with small specs of voxels that have been classified as the opposite orientation as the surrounding voxels. (C) Same as (B) but with median filtration applied. Examples of the trabecular orientation algorithm: (D) 36-year-old woman, (E) 66-year-old woman, (F) 83-year-old woman, (G) 32-year-old man, (H) 66-year-old man, and (I) 90-year-old man.
(p = 1.9 × 10−9) and men (p = 0.010). This shows that the absolute loss of bone volume is significantly higher for vertical than for horizontal bone volume. However, the balance between horizontal and vertical bone volume did not change significantly with age for women, whereas horizontal bone volume is lost relatively faster than vertical bone volume with age (r = −0.51, p = 9.7 × 10−4) for men (Fig. 3C). This indicates that the relative loss of vertical and horizontal bone volume is similar in women, whereas relatively more horizontal than vertical bone volume is lost with age in men. BV.horz/BV.vert was highly correlated with DA for both women (r = − 0.91, p = 4.4 × 10− 16) and for men (r = − 0.90, p = 4.0 × 10− 15). The thickness of vertical and horizontal trabeculae were independent of age for both women and men, and the regression coefficients
did not differ significantly between women and men for either relationship (Figs. 3D and E). However, this lack of statistical significance may be caused by relatively large inter individual variation in trabecular thickness. Thus, if the average thickness of the horizontal trabeculae was divided by the average thickness of the vertical trabeculae for each individual this ratio decreased significantly with age for both women (r = −0.54, p = 3.4 × 10−4) and men (r = −0.41, p = 0.0095) (Fig. 3F). This indicates that relatively more horizontal than vertical trabecular thickness is lost with age in both sexes. However, the slope of the fit line for Tb.Th*.vert differed significantly (p = 0.027) from that for Tb.Th*.horz, but for women only. This indicates that the absolute loss of trabecular thickness is more pronounced for horizontal trabeculae for women, whereas the absolute loss of horizontal and vertical trabecular thickness is similar in men.
J.S. Thomsen et al. / Bone 57 (2013) 47–55
51
Table 1 Regression coefficients (a, b) and Pearson's correlations coefficients (r) for the linear relationship f(x) = ax + b, where x is age and f(x) the different histomorphometric measures for women (n = 40), men (n = 39), and women and men combined (n = 79). An asterisk denotes a significant difference (p b 0.05) between the respective regression coefficient for women and men. A dagger denotes a significant difference between the respective regression coefficients for horizontal and vertical trabeculae. If no difference was found between the slope of the regression line for women and men an ANCOVA was performed. If no differences were found in any of the regression coefficients between women and men the regression analysis was also carried out for women and men combined. Women
ANCOVA
Women and men combined
a
b
r
p
a
Men b
r
p
p
a
b
r
p
BV/TV Tb.Th* Tb.Th SMI Tb.N* CD DA
−0.125 0.134 −0.165 0.0146* −0.00623 −0.0390 5.13 × 10−4*
17.0 120 104 0.645 1.38 6.57 1.45*
−0.83 0.15 −0.30 0.66 −0.77 −0.46 0.094
3.2 × 10−11 n.s. n.s. 3.5 × 10−6 5.6 × 10−9 0.0027 n.s.
−0.102 −0.0775 −0.161 0.00730 −0.00456 −0.0294 0.00272
16.6 141 110 0.937 1.28 5.80 1.31
−0.74 −0.066 −0.27 0.44 −0.63 −0.40 0.52
8.8 × 10−8 n.s. n.s. 0.0054 1.5 × 10−5 0.011 7.7 × 10−4
n.s n.s 0.0349 – n.s n.s –
−0.112 0.0335 −0.156 – −0.00536 −0.0342 –
16.7 130 107 – 1.33 6.19 –
−0.77 0.031 −0.26 – −0.70 −0.43 –
1.1 × 10−16 n.s. 0.019 – 5.3 × 10−13 6.5 × 10−5 –
BV.vert/TV BV.horz/TV BV.horz/BV.vert
−0.0874*,† −0.0373 −4.33 × 10−4*
12.2† 4.81 0.397*
−0.82 −0.75 −0.11
1.4 × 10−10 2.5 × 10−8 n.s.
−0.0600† −0.0417 −0.00231
11.2† 5.44 0.521
−0.68 −0.71 −0.51
2.1 × 10−6 3.4 × 10−7 9.7 × 10−4
– n.s. –
– −0.0392 –
– 5.11 –
– −0.71 –
– 1.7 × 10−13 –
Tb.Th*.vert Tb.Th*.horz Tb.Th*.horz/Tb.Th*.vert
0.210† −0.101 −0.00214
119 123 1.02
0.22 −0.12 −0.54
n.s. n.s. 3.4 × 10−4
−0.0338 −0.226 −0.00124
141 141 0.991
−0.028 −0.20 −0.41
n.s. n.s. 0.0095
n.s. 0.0223 n.s.
0.0920 −0.154 −0.00164
130 131 1.00
0.082 −0.15 −0.47
n.s. n.s. 1.5 × 10−5
Tb.Th*.vert.50th Tb.Th*.horz.50th Tb.Th*.vert.95th Tb.Th*.horz.95th
0.0174 −0.253 1.19 0.524
122 121 163 184
0.018 −0.30 0.47 0.21
n.s. n.s. 0.0021 n.s.
−0.110 −0.321 0.384 0.242
134 132 238 229
−0.11 −0.30 0.098 0.07
n.s. n.s. n.s. n.s.
n.s. n.s. 0.0432 0.0453
−0.0439 −0.280 0.800 0.410
128 126 199 205
−0.044 −0.29 0.23 0.13
n.s. 0.010 0.038 n.s.
Trabecular thickness distribution The 50th percentile (i.e. the median) of the trabecular thickness distribution did not change significantly with age for vertical or horizontal trabeculae for either women or men (Figs. 4A & B). The 95th percentile for the vertical trabecular thickness distributions increased significantly with age for women (r = 0.47, p=0.0021), but not for men (Fig. 4C). Actually, the 90th to the 99th percentiles for the vertical trabecular thickness distributions all increased significantly with age for women, but not for men. Moreover, the slope of the age-percentile fit lines increased monotonically from 0.88 μm/year (for the 90th percentile) to 1.46 μm/year (for the 99th percentile) for women, indicating a more pronounced increase in trabecular thickness with age for the thickest (highest percentile) vertical trabeculae. In contrast, the corresponding percentiles for the horizontal thickness distributions did not change significantly with age for either women or men (the 95th percentiles are depicted in Fig. 4D). Fig. 5 shows representative examples of the vertical thickness distribution for a 22-year-old woman (A), a 96-year-old woman (B), a 22year-old man (C), and a 90-year-old man (D). These vertical trabecular thickness distributions illustrate how the 95th percentile increased with age for the women, but not for the men. At the same time the median was similar for all four individuals, thus illustrating how the vertical thickness distribution changed shape with age for women, while remained almost unchanged with age for men. Discussion We here present a method for segmenting a set of 3D bone data into either vertical or horizontal voxels depending on the orientation of the bone in the local neighborhood and applied it to 79 human vertebral body bone specimens. In the present study we showed that the average trabecular bone volume fraction decreased significantly with age for both women and men, and that this age-related loss of bone density did not differ significantly between the two sexes. This is consistent with our previous findings for a subpopulation using 2D histomorphometry [26]. However, when the bone voxels were divided into horizontal and vertical voxels, we found that vertical bone was lost significantly faster with age for women than for men, whereas no sex-related differences were seen
for horizontal bone. In addition, the absolute loss of bone density was significantly larger for vertical oriented bone than for horizontal oriented bone for both sexes. In contrast, the relative loss of bone volume was similar for horizontal and vertical bone volume for women, whereas relatively more horizontal than vertical oriented bone volume was lost with age in men. The difference between the absolute and the relative bone losses is not surprising as the amount of vertical oriented bone volume was more than twice that of horizontal oriented bone volume. The present study showed that the average trabecular thickness determined with the direct method (i.e. assumption independent) Tb.Th* did not change significantly with age for either women or men. This is in accordance with the findings of Stauber and Müller, who investigated vertebral bone from 29 women and 38 men using μCT [8], and Liu et al. who investigated vertebral bone from 32 men using SEM [41]. However, Liu et al. did find a significant decrease in trabecular thickness from 20–29 years old men to 30–39 years old men, whereas no further decrease in trabecular thickness in older men was found. Furthermore, in the present study, we found that trabecular thickness determined using the parallel plate model Tb.Th decreased significantly with age for women and men when pooled. This is consistent with our previous finding using 2D histomorphometry, where we also found that Tb.Th decreased significantly with age when women and men were pooled [26]. The reason for this discrepancy between the findings for Tb.Th* and Tb.Th is most likely that the parallel plate model determines Tb.Th as 2 divided by the surface volume density, SV. With age the trabeculae are gradually converted from plate-like to rod-like (i.e. SMI increases with age), which will lead to a higher trabecular surface area per bone volume, i.e. a higher SV can be achieved with an unaltered trabecular thickness. As a consequence the trabecular thickness determined with the parallel plate model will increasingly be underestimated with age [42]. Although neither horizontal nor vertical trabecular thickness changed significantly with age, the horizontal to vertical trabecular thickness ratio decreased significantly with age for both sexes indicating that relatively more horizontal trabecular thickness is lost with age. However, the absolute trabecular thickness decreased significantly faster with age for horizontal trabeculae than for vertical trabeculae for women, whereas the loss of absolute trabecular thickness with age did not differ between horizontal and vertical trabeculae for men. It is interesting to note that the more rapid loss of vertical bone density in women than in men is not reflected in a similar more rapid loss of
52
J.S. Thomsen et al. / Bone 57 (2013) 47–55
12
225
A
D
200
10
Tb.Th*.vert (µm)
BV.vert/TV (%)
175 8 6 4
150 125 100 75 50
2 25 0
0 10 20 30 40 50 60 70 80 90 100
10 20 30 40 50 60 70 80 90 100
Age (years)
Age (years)
12
225
B
E
200
Tb.Th*.horz (µm)
BV.horz/TV (%)
10 8 6 4
175 150 125 100 75 50
2 25 0
0 10 20 30 40 50 60 70 80 90 100
10 20 30 40 50 60 70 80 90 100
Age (years)
Age (years)
1.0
1.2
BV.horz/BV.vert
F Tb.Th*.horz/Tb.Th*.vert
C 0.8
0.6
0.4
0.2
0.0
1.0
0.7
0.5
0.2
0.0 10 20 30 40 50 60 70 80 90 100
10 20 30 40 50 60 70 80 90 100
Age (years)
Age (years)
Fig. 3. Age-related changes of horizontal and vertical trabeculae. (A) Vertical trabecular bone volume. (B) Horizontal trabecular bone volume. (C) Ratio between horizontal and vertical bone volume. (D) Vertical trabecular thickness. (E) Horizontal trabecular thickness. (F) Ratio between horizontal and vertical trabecular thickness. Women are shown with open circles and dashed fit-lines, while males are shown with filled circles and solid fit-lines. Regression coefficients, correlation coefficients, and p-values for the linear relationships are shown in Table 1.
trabecular thickness in women than in men. The reason for this apparent paradox can be found in the trabecular bone structure. Either more vertical trabeculae are lost with age in women than in men, or more vertical plate-like trabeculae are converted into rod-like trabeculae in women than in men in order to account for the more pronounced loss of bone density in women than in men. The latter explanation is consistent with our findings for SMI, which showed that the conversion of plate-like trabeculae into rod-like trabeculae occurs significantly faster with age in women than in men. Macdonald et al. suggested that trabecular thinning might be a more common mechanism of bone loss in men owing to decreased rates of bone formation, whereas loss of trabecular elements might be more common in women owing to menopause-associated increases in osteoclastic activity leading to removal of entire trabeculae [43]. In the present study the loss of trabecular thickness with age was more pronounced in men than in women, although this was not statistically significant. Likewise, the loss of trabecular elements as assessed by Tb.N* was more pronounced in
women than in men, although this was borderline significant only. However, it should be born in mind that both SMI and Tb.N* were determined for the entire trabecular network and not for vertical trabeculae only. Further analysis of the trabecular thickness distributions showed that the thickest 5% of the vertical trabeculae (the 95th percentile) became significantly thicker with age for women, but not for men. Moreover, the age-related shift in the vertical trabecular thickness distribution towards thicker trabeculae in women was even more pronounced for the highest percentiles. At the same time the median and the mean trabecular thickness of both horizontal and vertical trabeculae was independent of age for both sexes. This indicates that the thickest vertical trabeculae underwent a compensatory thickening with age in women, but not in men. As mentioned above, we found that Tb.N* decreased borderline significantly faster in women than in men, indicating that more trabecular element may be lost with age in women than in men. This could explain why the compensatory hypertrophy is seen in
J.S. Thomsen et al. / Bone 57 (2013) 47–55
700
A
C
600
250
Tb.Th*.vert.95th (µm)
Tb.Th*.vert.50th (µm)
300
53
200 150 100 50
500 400 300 200 100
0
0 10 20 30 40 50 60 70 80 90 100
10 20 30 40 50 60 70 80 90 100
Age (years)
Age (years) 700
B
D
600
250
Tb.Th*.horz.95th (µm)
Tb.Th*.horz.50th (µm)
300
200 150 100 50
500 400 300 200 100
0
0 10 20 30 40 50 60 70 80 90 100
10 20 30 40 50 60 70 80 90 100
Age (years)
Age (years)
Fig. 4. Age related changes in 50th and 95th percentile for the horizontal and vertical trabecular thickness distribution. (A) 50th percentile (median) for the vertical trabeculae. (B) 50th percentile (median) for the horizontal trabeculae. (C) 95th percentile for the vertical trabeculae. (D) 95th percentile for the horizontal trabeculae. Women are shown with open circles and dashed fit-lines, while males are shown with filled circles and solid fit-lines. Regression coefficients, correlation coefficients and p-values for the linear relationships are shown in Table 1.
20
20
A
C 15
Frequency (%)
Frequency (%)
15
10
50th 95th
5
10
50th
5
0
95th
0 0
100
200
300
400
500
600
0
100
Tb.Th*.vert (µm)
200
300
400
500
20
20
B
D 15
Frequency (%)
15
Frequency (%)
600
Tb.Th*.vert (µm)
10
50th
5
95th
10
50th
5
0
95th
0 0
100
200
300
400
Tb.Th*.vert (µm)
500
600
0
100
200
300
400
500
600
Tb.Th*.vert (µm)
Fig. 5. Frequency distribution of Tb.Th*.vert for (A) a 22-year-old woman, (B) a 96-year-old woman, (C) a 22-year-old man, and (D) a 90-year-old man. The 50th percentile (median) is indicated with a dashed line and the 95th percentile is indicated with a solid line.
54
J.S. Thomsen et al. / Bone 57 (2013) 47–55
women but not in men, as fewer trabeculae are responsible for carrying the load in women with age leading to a higher load on the individual vertical trabeculae. Frost invited other researchers to provide proof that vertical trabeculae can strengthen and ‘thicken’ in adults under increased mechanical loading [18]. The present study provides an initial proof for this, as we demonstrated that the thickest vertical trabeculae became thicker with age, however, in women only. Nevertheless, without longitudinal data it is difficult to establish with certainty whether the vertical trabeculae actually thicken or there is a disproportional removal of thinner trabeculae resulting in the observed changes in the vertical thickness distribution. Whether these morphological differences between women and men influence fracture risk is not clear and merit further investigations. Bergot et al. analyzed vertebral bone from 31 women and 30 men at various ages and determined the mean trabecular thickness of both horizontal and vertical trabeculae using 2D granulometry [21]. They found trabecular thickness distributions similar to those observed in the present study, although the peak in the thickness distribution plot was higher than that observed in the present study probably due to a larger bin width. In contrast to the present study, they observed a slight shift to the left in the vertical thickness distribution histograms with age for both women and men. This discrepancy may be due to that the study by Bergot et al. was based on 2D methods, whereas the present study employed 3D methods. The method described in the present study is an extension of our previously introduced method for analyzing 2D histological sections [24]. A similar approach was taken by Mc Donnell et al., who investigated three vertebral bodies in 3D at a spatial resolution of 30–36 μm [13]. However, in the present work we found that for high resolution scans (in this case a spatial resolution of 18.5 μm) the original algorithm identifies tiny protrusions on the trabeculae as trabeculae of the orthogonal direction. We showed that these erroneously identified voxels could be reclassified without compromising the overall identification of the trabeculae by a simple median filtration. The presented method is voxel based i.e. the trabecular network is subdivided into horizontal and vertical oriented bone tissue rather than horizontal and vertical oriented trabeculae, although in most cases the outcome would be similar. Stauber and Müller [8,44–46] and Liu et al. [47–49] have both presented a method for subdividing the trabecular network into individual trabeculae, thus making it possible to classify each trabecula as a plate or a rod and to quantify its spatial orientation. Consequently, it would be interesting to compare the capability of our relatively simple voxel based method to that of the more complex trabecular based method of Liu et al. and Stauber and Müller in a future study. In addition, with slight modifications of the scanner software other measures such as e.g. SMI can also be determined for horizontal and vertical trabeculae separately using the method described in the present paper. We find that the presented method is easy to implement and can, at least for Scanco μCT scanners, be applied retrospectively, and – based on visual inspection – provides convincing results. A limitation of the present study is that we investigated an approximately 9-mm-thick bone slice from the central part of the vertebrae. We have previously shown that the trabecular bone volume and micro architecture is not homogeneously distributed in the vertical direction in human vertebral bodies [50]. Likewise, some studies have indicated that there are differences in the distribution of bone density and micro structure between the anterior and posterior regions of the vertebra [51–53]. In the present study we used a fixed global threshold filtration procedure for segmentation of the scans in accordance with the recommendations given in the current μCT guidelines [54]. Some researchers argue that application of more complex segmentation methods can be advantageous [55–57]. However, since no differences in the agerelated changes between women and men in trabecular bone ashdensity were found in the adjacent L3 [25], there is no reason to believe that the applied segmentation technique influenced the sex-related differences found in the present study. Finally, the partial volume effect,
which is an inherent limitation to the CT technique, could potentially bias the measurements although this limitation is most pronounced for voxels sizes larger than that used in the present study [54]. Conclusion In the present study we have described an easy method for segmenting a trabecular network into horizontal and vertical oriented bone. Using this method on 79 human lumbar bone samples, we have shown that vertical and horizontal bone decreased with age in both women and men, and that vertical oriented bone was lost more rapidly in women than in men. Furthermore, vertical as well as horizontal trabecular thickness were found to be independent of age, whereas the horizontal/vertical trabecular thickness ratio decreased significantly with age indicating a more pronounced thinning of horizontal trabeculae. Finally, we found indications that the age-related loss of trabecular elements resulted in compensatory hypertrophy of vertical trabeculae in women, but not in men. Disclosures All authors state that they have no conflicts of interest. Acknowledgments Andres Laib and Stephan Weiss, Scanco Medical AG, are gratefully acknowledged for help with IPL and decoding Scanco's AIM image file format. The μCT scanner was donated by the VELUX Foundation. The Helga and Peter Korning Foundation and the A.P. Møller Foundation for the Advancement of Medical Science are acknowledged for financial support of the study. Authors' roles: Conception and design: JST and AB. Acquisition of data: JST, ASN, and ENE. Analysis and interpretation of data: JST, ASN, ENE, AB. Drafting manuscript: JST. Revising manuscript: JST, ASN, ENE, and AB. Approved final version of manuscript: JST, ASN, ENE, AB. Responsibility for integrity of data analysis: JST. References [1] Ebbesen EN, Thomsen JS, Beck-Nielsen H, Nepper-Rasmussen HJ, Mosekilde Li. Lumbar vertebral body compressive strength evaluated by dual-energy X-ray absorptiometry, quantitative computed tomography, and ashing. Bone 1999;25:713–24. [2] Perilli E, Briggs AM, Kantor S, Codrington J, Wark JD, Parkinson IH, et al. Failure strength of human vertebrae: prediction using bone mineral density measured by DXA and bone volume by micro-CT. Bone 2012;50:1416–25. [3] Kleerekoper M, Villanueva AR, Stanciu J, Rao DS, Parfitt AM. The role of three-dimensional trabecular microstructure in the pathogenesis of vertebral compression fractures. Calcif Tissue Int 1985;37:594–7. [4] Dempster DW. Bone microarchitecture and strength. Osteoporos Int 2003;14(Suppl. 5):S54–6. [5] Räth C, Monetti R, Bauer J, Sidorenko I, Müller D, Matsuura M, et al. Strength through structure: visualization and local assessment of the trabecular bone structure. New J Phys 2008;10:125010. [6] Parkinson IH, Badiei A, Stauber M, Codrington J, Müller R, Fazzalari NL. Vertebral body bone strength: the contribution of individual trabecular element morphology. Osteoporos Int 2012;23:1957–65. [7] 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. [8] Stauber M, Müller R. Age-related changes in trabecular bone microstructures: global and local morphometry. Osteoporos Int 2006;17:616–26. [9] Bell GH, Dunbar O, Beck JS, Gibb A. Variations in strength of vertebrae with age and their relation to osteoporosis. Calcif Tissue Res 1967;1:75–86. [10] Jensen KS, Li Mosekilde, Le Mosekilde. A model of vertebral trabecular bone architecture and its mechanical properties. Bone 1990;11:417–23. [11] Li Mosekilde. Sex differences in age-related loss of vertebral trabecular bone mass and structure—biomechanical consequences. Bone 1989;10:425–32. [12] Atkinson PJ. Variation in trabecular structure of vertebrae with age. Calcif Tissue Res 1967;1:24–32. [13] Mc Donnell P, Harrison N, Liebschner MA, Mc Hugh PE. Simulation of vertebral trabecular bone loss using voxel finite element analysis. J Biomech 2009;42:2789–96. [14] van der Linden JC, Homminga J, Verhaar JA, Weinans H. Mechanical consequences of bone loss in cancellous bone. J Bone Miner Res 2001;16:457–65.
J.S. Thomsen et al. / Bone 57 (2013) 47–55 [15] Fields AJ, Lee GL, Liu XS, Jekir MG, Guo XE, Keaveny TM. Influence of vertical trabeculae on the compressive strength of the human vertebra. J Bone Miner Res 2011;26: 263–9. [16] Homminga J, Van-Rietbergen B, Lochmüller 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. [17] Fields AJ, Nawathe S, Eswaran SK, Jekir MG, Adams MF, Papadopoulos P, et al. Vertebral fragility and structural redundancy. J Bone Miner Res 2012;27:2152–8. [18] Frost HM. On the trabecular “thickness”-number problem. J Bone Miner Res 1999;14:1816–21. [19] Casuccio C. An introduction to the study of osteoporosis (biochemical and biophysical research in bone ageing). Proc R Soc Med 1962;55:663–8. [20] Henschke F, Pesch HJ. Spongiosastruktur von lendenwirbelkörpern und schenkelhals vergleichende analyse des altersabhängigen umbaues. Microsc Acta Suppl 1980;4:124–8. [21] Bergot C, Laval-Jeantet AM, Prêteux F, Meunier A. Measurement of anisotropic vertebral trabecular bone loss during aging by quantitative image analysis. Calcif Tissue Int 1988;43:143–9. [22] Jayasinghe JAP, Jones SJ, Boyde A. Three-dimensional photographic study of cancellous bone in human fourth lumbar vertebral bodies. Anat Embryol (Berl) 1994;189: 259–74. [23] Parfitt AM, Mathews CHE, 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 cellular mechanisms of bone loss. J Clin Invest 1983;72:1396–409. [24] 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. [25] Ebbesen EN, Thomsen JS, Beck-Nielsen H, Nepper-Rasmussen HJ, Mosekilde Li. Ageand gender-related differences in vertebral bone mass, density, and strength. J Bone Miner Res 1999;14:1394–403. [26] Thomsen JS, Ebbesen EN, Mosekilde Li. Static histomorphometry of human iliac crest and vertebral trabecular bone: a comparative study. Bone 2002;30:267–74. [27] Ebbesen EN, Thomsen JS, Beck-Nielsen H, Nepper-Rasmussen HJ, Mosekilde Li. Vertebral bone density evaluated by dual-energy X-ray absorptiometry and quantitative computed tomography in vitro. Bone 1998;23:283–90. [28] Thomsen JS, Ebbesen EN, Mosekilde Li. A new method of comprehensive static histomorphometry applied on human lumbar vertebral cancellous bone. Bone 2000;27:129–38. [29] Sikjaer T, Rejnmark L, Thomsen JS, Tietze A, Brüel A, Andersen G, et al. Changes in 3-dimensional bone structure indices in hypoparathyroid patients treated with PTH(1–84): a randomized controlled study. J Bone Miner Res 2012;27:781–8. [30] Hildebrand T, Laib A, Müller R, Dequeker J, Rügsegger P. Direct three-dimensional morphometric analysis of human cancellous bone: microstructural data from spine, femur, iliac crest, and calcaneus. J Bone Miner Res 1999;14:1167–74. [31] Hildebrand T, Rügsegger P. A new method for the model-independent assessment of thickness in three-dimensional images. J Microsc 1997;185:67–75. [32] Laib A, Rüegsegger P. Comparison of structure extraction methods for in vivo trabecular bone measurements. Comput Med Imaging Graph 1999;23:69–74. [33] Odgaard A, Gundersen HJG. Quantification of connectivity in cancellous bone, with special emphasis on 3-D reconstructions. Bone 1993;14:173–82. [34] Hildebrand T, Rügsegger P. Quantification of bone microarchitecture with the structure model index. Comput Methods Biomech Biomed Engin 1997;1:15–23. [35] Thomsen JS, Laib A, Koller B, Prohaska S, Mosekilde Li, Gowin W. Stereological measures of trabecular bone structure: comparison of 3D micro computed tomography with 2D histological sections in human proximal tibial bone biopsies. J Microsc 2005;218:171–9.
55
[36] Laib A, Barou O, Vico L, Lafage-Proust MH, Alexandre C, Rügsegger P. 3D microcomputed tomography of trabecular and cortical bone architecture with application to a rat model of immobilisation osteoporosis. Med Biol Eng Comput 2000;38:326–32. [37] Bresenham JE. Algorithm for computer control of a digital plotter. IBM Syst J 1965;4: 25–30. [38] Thomsen JS, Ebbesen EN, Mosekilde Li. Relationships between static histomorphometry and bone strength measurements in human iliac crest bone biopsies. Bone 1998;22:153–63. [39] Diem K, Lentner C. Scientific tables. 7th ed. Basle, J.R: Geigy S.A; 1970 . [40] Press WH, Teukolsky SA, Vetterling WT, Flannery BP. Numerical recipes in C — the art of scientific computing. Cambridge: Cambridge University Press; 1992 . [41] Liu GM, Xu CJ, Kong N, Zhu XM, Zhang XY, Yao Y. Age-related differences in microstructure, density and biomechanics of vertebral cancellous bone of Chinese males. Aging Male 2012;15:233–9. [42] Day JS, Ding M, Odgaard A, Sumner DR, Hvid I, Weinans H. Parallel plate model for trabecular bone exhibits volume fraction-dependent bias. Bone 2000;27:715–20. [43] Macdonald HM, Nishiyama KK, Kang J, Hanley DA, Boyd SK. Age-related patterns of trabecular and cortical bone loss differ between sexes and skeletal sites: a population-based HR-pQCT study. J Bone Miner Res 2011;26:50–62. [44] Stauber M, Müller R. Volumetric spatial decomposition of trabecular bone into rods and plates—a new method for local bone morphometry. Bone 2006;38:475–84. [45] Stauber M, Rapillard L, van Lenthe GH, Zysset P, Müller R. Importance of individual rods and plates in the assessment of bone quality and their contribution to bone stiffness. J Bone Miner Res 2006;21:586–95. [46] Stauber M, Müller R. A sensitivity analysis of the volumetric spatial decomposition algorithm. Comput Methods Biomech Biomed Engin 2007;10:25–37. [47] Liu XS, Sajda P, Saha PK, Wehrli FW, Guo XE. Quantification of the roles of trabecular microarchitecture and trabecular type in determining the elastic modulus of human trabecular bone. J Bone Miner Res 2006;21:1608–17. [48] Liu XS, Sajda P, Saha PK, Wehrli FW, Bevill G, Keaveny TM, et al. Complete volumetric decomposition of individual trabecular plates and rods and its morphological correlations with anisotropic elastic moduli in human trabecular bone. J Bone Miner Res 2008;23:223–35. [49] Liu XS, Zhang XH, Guo XE. Contributions of trabecular rods of various orientations in determining the elastic properties of human vertebral trabecular bone. Bone 2009;45:158–63. [50] Thomsen JS, Ebbesen EN, Mosekilde Li. Zone-dependent changes in human vertebral trabecular bone: clinical implications. Bone 2002;30:664–9. [51] Sandor T, Felsenberg D, Kalender WA, Brown E. Global and regional variations in the spinal trabecular bone: single and dual energy examinations. J Clin Endocrinol Metab 1991;72:1157–68. [52] 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. [53] Hussein AI, Morgan EF. The effect of intravertebral heterogeneity in microstructure on vertebral strength and failure patterns. Osteoporos Int 2013;24:979–89. [54] Bouxsein ML, Boyd SK, Christiansen BA, Guldberg RE, Jepsen KJ, Müller R. Guidelines for assessment of bone microstructure in rodents using micro-computed tomography. J Bone Miner Res 2010;25:1468–86. [55] Scherf H, Tilgner R. A new high-resolution computed tomography (CT) segmentation method for trabecular bone architectural analysis. Am J Phys Anthropol 2009;140:39–51. [56] Burghardt AJ, Kazakia GJ, Majumdar S. A local adaptive threshold strategy for high resolution peripheral quantitative computed tomography of trabecular bone. Ann Biomed Eng 2007;35:1678–86. [57] Dufresne T. Segmentation techniques for analysis of bone by three-dimensional computed tomographic imaging. Technol Health Care 1998;6:351–9.
Contents lists available at ScienceDirect
Bone journal homepage: www.elsevier.com/locate/bone
Technical Note
Age-related changes of vertical and horizontal lumbar vertebral trabecular 3D bone microstructure is different in women and men Jesper Skovhus Thomsen ⁎, Andreas Steenholt Niklassen, Ebbe Nils Ebbesen, Annemarie Brüel Department of Biomedicine — Anatomy, Aarhus University, Aarhus, Denmark
a r t i c l e
i n f o
Article history: Received 9 April 2013 Revised 15 July 2013 Accepted 16 July 2013 Available online 27 July 2013 Edited by: David Burr Keywords: Bone μCT Bone structure Static histomorphometry Human vertebra Sex differences Aging
a b s t r a c t The study presents a 3D method for subdividing a trabecular network into horizontal and vertical oriented bone. This method was used to investigate the age related changes of the bone volume fraction and thickness of horizontal and vertical trabeculae in human lumbar vertebral bone estimated with unbiased 3D methods in women and men over a large age-range. The study comprised second lumbar vertebral body bone samples from 40 women (aged 21.7–96.4 years, median 56.6 years) and 39 men (aged 22.6–94.6 years, median 55.6 years). The bone samples were μCT scanned and the 3D microstructure was quantified. A voxel based algorithm inspecting the local neighborhood is presented and used to segment the trabecular network into horizontal and vertical oriented bone. For both women and men BV/TV decreased significantly with age, Tb.Th* was independent of age, while SMI increased significantly with age. Vertical (BV.vert/TV) and horizontal (BV.horz/TV) bone volume fraction decreased significantly with age for both sexes. BV.vert/TV decreased significantly faster with age for women than for men. Vertical (Tb.Th*.vert) and horizontal (Tb.Th*.horz) trabecular thickness were independent of age, while Tb.Th*.horz/Tb.Th*.vert decreased significantly with age for both sexes. Additionally, the 95th percentile of the trabecular thickness distribution increased significantly with age for vertical trabeculae in women, whereas it was independent of age in men. In conclusion, we have shown that vertical and horizontal oriented bone density decreases with age in both women and men, and that vertical oriented bone is lost more quickly in women than in men. Furthermore, vertical and horizontal trabecular thickness were independent of age, whereas the horizontal to vertical trabecular thickness ratio decreased significantly with age indicating a relatively more pronounced thinning of horizontal trabeculae. Finally, the age-related loss of trabecular elements appeared to result in a compensatory hypertrophy of vertical trabeculae in women, but not in men. © 2013 Elsevier Inc. All rights reserved.
Introduction During aging, human vertebral bone is lost resulting in weaker bone and thereby a higher fracture risk. It has been shown that bone density is a major determinant of vertebral bone strength [1,2], but it has also been suggested that the microstructure of the bone tissue plays a role for the bone strength [3–6]. Hence, Hui et al. showed that even for constant bone mass, fracture risk increases with age [7]. Thus, bone fracture strength is not only dependent on bone density, but also on bone microstructure, micro damage accumulation, and mineralization [8]. The conventional view is that a compressive load on a vertebral body is mainly carried by the vertical trabeculae, whereas the horizontal trabeculae serve to prevent buckling of the vertical trabeculae [9–12]. This view is reinforced by finite element analyses of human vertebral bone ⁎ Corresponding author at: Department of Biomedicine — Anatomy, Aarhus University, Wilhelm Meyers Allé 3, DK-8000 Aarhus C, Denmark. E-mail addresses: [email protected] (J.S. Thomsen), [email protected] (A.S. Niklassen), [email protected] (E.N. Ebbesen), [email protected] (A. Brüel). 8756-3282/$ – see front matter © 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.bone.2013.07.025
specimens, which demonstrates that vertical trabeculae are more highly strained than horizontal trabeculae under normal compressive loading [13–16]. Furthermore, Fields et al. hypothesized that vertebral bone strength is better explained by the bone volume fraction of the vertical trabeculae alone, than by the bone volume fraction of all trabeculae [15]. In a subsequent study, Fields et al. found that the vertical trabeculae played a particular important role for the compressive bone strength of vertebrae with low bone density [17]. Consequently, it is important to quantify the age-related changes in trabecular thickness as well as bone volume fraction for horizontal and vertical trabeculae separately. It has, for a long time, been debated whether the removal of vertical trabeculae with age would lead to a higher compressive load on the remaining trabeculae, and thereby to a compensatory thickening of these [12,18–22]. Thus, in 1983 Parfitt et al. suggested that, when trabeculae are removed during aging, the remaining trabeculae would be more widely separated and therefore might undergo compensatory thickening [23]. Furthermore, in a 1999 perspective article Harold Frost invited other researchers to provide proof that vertical trabeculae can strengthen and ‘thicken’ in adults under increased mechanical
48
J.S. Thomsen et al. / Bone 57 (2013) 47–55
loading [18]. We have previously investigated this using 2D histomorphometry on human vertebral bone, but were unable to find evidence of a compensatory thickening of vertebral trabeculae with age [24]. However, the 2D analysis was limited by the use of the parallel plate model for determination of the trabecular thickness. The development of micro computed tomography (μCT) scanners has now made it feasible to investigate the microstructure of the 3D trabecular network in human vertebrae using model assumption free methods for determination of the trabecular thickness. Therefore, the aim of the present study was to present a 3D method for segmenting the trabecular network into horizontal and vertical oriented bone. Furthermore, we applied the method to 79 human vertebral bone specimens and tested the hypothesis that the age related changes of the bone volume fraction and thickness of horizontal and vertical trabeculae in human vertebral bone do not differ in women and men. Finally, we investigated whether vertical vertebral trabeculae can thicken with age. Materials and methods Bone specimens Second lumbar vertebral bodies from 40 women (aged 21.7– 96.4 years, median 56.6 years) and 39 men (aged 22.6–94.6 years, median 55.6 years) were analyzed. The subjects were chosen so as to provide an even age and sex distribution in the 20–90 years age range. All individuals were Caucasian. Vertebral bodies with fractures identified from lateral-projection X-ray images were excluded from further investigations as previously reported [1]. Samples from individuals with known cancer or drug abuse were excluded from the study at the time of autopsy. Subsequently, a thorough review of hospital records and autopsy information available was used to exclude individuals with cancer (discovered at autopsy), metabolic disease, severe liver or kidney disease, medication affecting bone metabolism, or periods of more than 2 weeks of immobilization prior to death as previously described [25]. The vertebral bodies from 24 of the women and 24 of the men have previously been investigated using 2D histomorphometry [24,26]. The investigation presented here is a subset of a larger study called the “Danish in Vitro Bone Study” (DAVIBO) [1,25,27]. The collection of the material and the study design were approved by the local ethical committee. Specimen preparation
was interactively delineated as closely as possible to either the cortical bone or the sawing planes using custom-made software written by one of the authors (JST) [29]. The resulting image masks were transferred back to the μCT scanner and imported into IPL (version 5.11, Scanco Medical). The 3D data sets were low-pass filtered using a Gaussian filter (σ = 0.8, support = 1) in order to remove noise and were subsequently segmented with a fixed threshold filter. The minimum point between the marrow and the bone peak in the attenuation histogram was automatically determined using IPL for 25 vertebral bone specimens, and the median of these thresholds (450.7 mg HA/cm3) was used for all segmentations. Standard quantification of the microstructure of the trabecular bone network was performed using the software provided with the μCT scanner (version 6.0, Scanco Medical). Microstructural measures included bone volume per total volume (BV/TV), trabecular thickness (parallel plate model: Tb.Th [30] and directly estimated: Tb.Th* [31]), trabecular number (Tb.N*) [32], connectivity density (CD) [33], and structural model index (SMI) [34]. The computation of these structural measures has previously been described in detail [35]. In addition, the degree of anisotropy (DA) was also computed [36]. Quality assurance was performed by weekly (density) and monthly (geometry) scans of the solid-state calibration phantom provided with the scanner. Trabecular orientation The voxels of the trabecular bone network was classified as either horizontal or vertical by inspecting the orientation of the bone tissue surrounding the voxel. The method is an extension of our previously presented 2D method into 3D [24]. Consider a voxel characterized by its Cartesian coordinates (x0, y0, z0). Let l*(θ, φ) denote the Euclidian length from the voxel at (x0, y0, z0) to either the bone-marrow boundary or the VOI boundary, where θ is the polar angle and φ is the azimuthal angle (Fig. 1). Let
hðx0 ; y0 ; z0 Þ ¼
N −1 X
3N 8 −1
X
j¼0 i¼N8
vðx0 ; y0 ; z0 Þ ¼
N N −1 X 8 −1 X
j¼0 i¼0
i j l 2π ; 2π N N
ð1Þ
NX −1 N −1 X i j i j þ l 2π ; 2π l 2π ; 2π N N N N 3N j¼0
ð2Þ
i¼ 8
The bone specimens had previously been prepared for histomorphometry [28]: The vertebral bodies were cleaned of soft connective tissue and halved along the anterior–posterior axis using a diamond parallel precision saw (Exakt, Apparatebau, Otto Herrmann, Norderstedt, Germany). An approximately 9-mm-thick frontal bone specimen was sawed from the center of one half of the corpus and embedded undecalcified in methyl methacrylate (Technovit 9100, Heraeus Kulzer, Wehrheim/Ts., Germany). Subsequently, the tissue blocks were trimmed using the diamond saw to remove all excessive methyl methacrylate to facilitate μCT scanning. Micro computed tomography (μCT) The bone specimens were placed in a μCT scanner (μCT35, Scanco Medical AG, Brüttisellen, Switzerland) so that they were aligned along the x-axis of the scanner and so the vertebral endplates were parallel with the xy-plane. The specimens were scanned in high-resolution mode (1000 projections per 180°) with an isotropic voxel size of 18.5 μm, an X-ray tube voltage of 70 kVp, an X-ray tube current of 114 μA, and an integration time of 800 ms. The reconstructed μCT data was exported to a PC running Linux (OpenSUSE 11.2, http://www. opensuse.org), where the trabecular bone volume of interest (VOI)
Fig. 1. Coordinate system with the origin placed in the voxel under inspection (x0, y0, z0) used for the calculations. The Euclidian length from the origin to either the bone-marrow boundary or the VOI boundary is denoted l*(θ, φ), where θ is the polar angle and φ is the azimuthal angle.
J.S. Thomsen et al. / Bone 57 (2013) 47–55
where N is a multiple of 8. If h(x0,y0,z0) N v(x0,y0,z0) the voxel at (x0, y0, z0) is classified as horizontal, otherwise it is classified as vertical. In all, l* is determined for N2/2 horizontal and N2/2 vertical directions for each voxel. In the present study N = 128 and consequently l* was determined for 1282 = 16,384 directions for each voxel. Only voxels representing bone inside the VOI is considered, while voxels representing marrow is not considered. In order to avoid single voxel protrusions on the trabeculae as being identified as a separate trabecula, the resulting data set underwent a median filtration. The standard software provided with the Scanco scanner outputs a file containing the “local” Tb.Th* values determined for all bone voxels. The standard software computes the Tb.Th* value for the entire trabecular network as the mean of these local Tb.Th* values. As these thickness values are determined using the direct estimation method they are not biased by model assumptions. Thus, by combining the h and v sets with the Tb.Th* file it is possible to estimate the average thickness of horizontal trabeculae, Tb.Th*.horz, and vertical trabeculae, Tb.Th*.vert, respectively. The h and v data sets were exported and saved for subsequent 3D visualization and further data processing. The above described method was implemented in a custom made computer program written in C and running under Linux using the Tb.Th* file as input. This file serves the multiple purpose of identifying bone voxels inside the VOI (a Tb.Th* value larger than 0) and for computation of Tb.Th*.horz and Tb.Th*.vert. As l*(θ, φ) has to be calculated many times for each voxel, the algorithm for determining l*(θ, φ) has to be very fast. We implemented a 3D version of Bresenham's line algorithm [37] for this purpose, which is normally considered very efficient. From the h and v data sets the horizontal bone volume fraction, BV.horz/TV, and the vertical bone volume fraction, BV.vert/TV, can easily be computed as the number of voxels in either the h or v data set divided by the total number of voxels in the VOI. In addition, the thickness distribution for both the horizontal and the vertical direction was determined and the 50th and the 90th–99th percentiles were determined for these distributions. 3D visualization The h and v data sets were 3D visualized as isosurfaces for selected bone samples using Amira 4.1.2 (Mercury Computer Systems, Inc., Chelmsford, Mass., USA) running under Linux. Statistics Linear regression was performed using the least-squares method using custom made software. Comparisons between two linear relationships y1 = a1 x + b1 and y2 = a2 x + b2 each representing N data points were carried out using a t-test [38]. For comparison between the two slopes, a1 and a2, the test statistic was [39]: ja1 −a2 j ffi t ¼ vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u X 2 X 2 uN y1 − y1 1−r 2 u 1 u t X 2 X 2 N−2 N x − x 1
ð3Þ
Both tests were carried out with ν = N − 2 degrees of freedom. The probability level p for the test statistic t was computed as [40]: t
∫
1þ
−t
pðt Þ ¼ 1−
1
ν
1=2
x2 ν
−1=2
∫x
!−ðνþ1Þ=2 dx ð5Þ ðν=2Þ−1
ð1−xÞ
dx
0
p b 0.05 was considered significant. If the slope of the regression line did not differ between women and men an analysis of covariance (ANCOVA) was performed using the statistical software R (http://www.r-project.org). Results Trabecular orientation Fig. 2A shows an example of the voxel orientation classification algorithm. Horizontal voxels are shown in blue and vertical voxels are shown in red. The efficacy of the algorithm can clearly be seen from the 3D visualization. Figs. 2B and C shows a magnification of the marked region in Fig. 2A. Fig. 2B is computed without median filtration, whereas Fig. 2C is computed with median filtration. It can clearly be seen that the small specs of blue horizontal voxels that can be observed in Fig. 2B, have been filtered out in Fig. 2C without compromising the shape and extent of the trabeculae. Figs. 2D–I show examples of the algorithm for women and men at different ages. Age-related changes in bone architecture Measurements of bone architecture are given in Table 1. BV/TV decreased significantly with age for both women (r = −0.83, p = 3.2 × 10−11) and men (r = −0.74, p = 8.8 × 10−8). These agerelated changes in BV/TV did not differ significantly between women and men. Average trabecular thickness (Tb.Th*) determined using the direct method did not change significantly with age for either women or men. Similarly, average trabecular thickness determined using parallel plate assumptions (Tb.Th) did also not decrease significantly with age for women or men. However, if women and men were pooled Tb.Th decreased significantly with age (r = −0.26, p = 0.019). SMI increased significantly with age for women (r = 0.66, p = 3.5 × 10−6) and men (r = 0.44, p = 0.0054). The fit line was significantly steeper (p = 0.0054) for women than for men, indicating a faster transformation of the trabecular bone network from plate-like to rod-like in women. Tb.N* decreased significantly with age for both women (r = −0.77, p = 5.6 × 10−9) and men (r = −0.63, p = 1.5 × 10−5), and likewise CD decreased significantly with age for both women (r = −0.46, p = 0.0027) and men (r = −0.40, p = 0.011). These age-related changes in Tb.N* and CD did not differ between the sexes. The degree of anisotropy (DA) was independent of age for women (r = 0.094, n.s.), whereas it increased significantly with age for men (r = 0.52, p = 7.7 × 10−4). Age-related changes of horizontal and vertical trabeculae
1
where r1 is Pearson's correlation coefficient. For comparison between the two y-axis intercepts, b1 and b2, the test statistic was [39]: jb1 −b2 j ffi t ¼ vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi uX X 2 X 2 u 2 2 N y − y u x1 1 1 1−r 1 u X 2 X 2 N−2 t N N x1 − x1
49
ð4Þ
The trabecular bone volume fraction of vertical voxels (BV.vert/TV) decreased significantly with age for women (r = −0.82, p = 1.4 × 10−10) and men (r = −0.68, p = 2.1 × 10−6) (Fig. 3A). The slope of the fit line differed significantly (p = 0.014) between women and men, showing that BV.vert/TV decreased significantly faster with age for women than for men. BV.horz/TV also decreased significantly with age for women (r = −0.75, p = 2.5 × 10−8) and men (r = −0.71, p = 3.4 × 10−7), but these age-related difference did not differ between the sexes (Fig. 3B). The fit line for BV.vert/TV was significantly steeper than the fit line for BV.horz/TV for both women
50
J.S. Thomsen et al. / Bone 57 (2013) 47–55
B
A
C
D
G
E
F
H
I
Fig. 2. 3D visualization of the trabecular network subdivided into horizontal (blue) and vertical (red) voxels. (A) 27-year-old man. (B) Magnification of the region shown with a box in (A) without median filtration (see text for detail). Arrowheads indicates regions with small specs of voxels that have been classified as the opposite orientation as the surrounding voxels. (C) Same as (B) but with median filtration applied. Examples of the trabecular orientation algorithm: (D) 36-year-old woman, (E) 66-year-old woman, (F) 83-year-old woman, (G) 32-year-old man, (H) 66-year-old man, and (I) 90-year-old man.
(p = 1.9 × 10−9) and men (p = 0.010). This shows that the absolute loss of bone volume is significantly higher for vertical than for horizontal bone volume. However, the balance between horizontal and vertical bone volume did not change significantly with age for women, whereas horizontal bone volume is lost relatively faster than vertical bone volume with age (r = −0.51, p = 9.7 × 10−4) for men (Fig. 3C). This indicates that the relative loss of vertical and horizontal bone volume is similar in women, whereas relatively more horizontal than vertical bone volume is lost with age in men. BV.horz/BV.vert was highly correlated with DA for both women (r = − 0.91, p = 4.4 × 10− 16) and for men (r = − 0.90, p = 4.0 × 10− 15). The thickness of vertical and horizontal trabeculae were independent of age for both women and men, and the regression coefficients
did not differ significantly between women and men for either relationship (Figs. 3D and E). However, this lack of statistical significance may be caused by relatively large inter individual variation in trabecular thickness. Thus, if the average thickness of the horizontal trabeculae was divided by the average thickness of the vertical trabeculae for each individual this ratio decreased significantly with age for both women (r = −0.54, p = 3.4 × 10−4) and men (r = −0.41, p = 0.0095) (Fig. 3F). This indicates that relatively more horizontal than vertical trabecular thickness is lost with age in both sexes. However, the slope of the fit line for Tb.Th*.vert differed significantly (p = 0.027) from that for Tb.Th*.horz, but for women only. This indicates that the absolute loss of trabecular thickness is more pronounced for horizontal trabeculae for women, whereas the absolute loss of horizontal and vertical trabecular thickness is similar in men.
J.S. Thomsen et al. / Bone 57 (2013) 47–55
51
Table 1 Regression coefficients (a, b) and Pearson's correlations coefficients (r) for the linear relationship f(x) = ax + b, where x is age and f(x) the different histomorphometric measures for women (n = 40), men (n = 39), and women and men combined (n = 79). An asterisk denotes a significant difference (p b 0.05) between the respective regression coefficient for women and men. A dagger denotes a significant difference between the respective regression coefficients for horizontal and vertical trabeculae. If no difference was found between the slope of the regression line for women and men an ANCOVA was performed. If no differences were found in any of the regression coefficients between women and men the regression analysis was also carried out for women and men combined. Women
ANCOVA
Women and men combined
a
b
r
p
a
Men b
r
p
p
a
b
r
p
BV/TV Tb.Th* Tb.Th SMI Tb.N* CD DA
−0.125 0.134 −0.165 0.0146* −0.00623 −0.0390 5.13 × 10−4*
17.0 120 104 0.645 1.38 6.57 1.45*
−0.83 0.15 −0.30 0.66 −0.77 −0.46 0.094
3.2 × 10−11 n.s. n.s. 3.5 × 10−6 5.6 × 10−9 0.0027 n.s.
−0.102 −0.0775 −0.161 0.00730 −0.00456 −0.0294 0.00272
16.6 141 110 0.937 1.28 5.80 1.31
−0.74 −0.066 −0.27 0.44 −0.63 −0.40 0.52
8.8 × 10−8 n.s. n.s. 0.0054 1.5 × 10−5 0.011 7.7 × 10−4
n.s n.s 0.0349 – n.s n.s –
−0.112 0.0335 −0.156 – −0.00536 −0.0342 –
16.7 130 107 – 1.33 6.19 –
−0.77 0.031 −0.26 – −0.70 −0.43 –
1.1 × 10−16 n.s. 0.019 – 5.3 × 10−13 6.5 × 10−5 –
BV.vert/TV BV.horz/TV BV.horz/BV.vert
−0.0874*,† −0.0373 −4.33 × 10−4*
12.2† 4.81 0.397*
−0.82 −0.75 −0.11
1.4 × 10−10 2.5 × 10−8 n.s.
−0.0600† −0.0417 −0.00231
11.2† 5.44 0.521
−0.68 −0.71 −0.51
2.1 × 10−6 3.4 × 10−7 9.7 × 10−4
– n.s. –
– −0.0392 –
– 5.11 –
– −0.71 –
– 1.7 × 10−13 –
Tb.Th*.vert Tb.Th*.horz Tb.Th*.horz/Tb.Th*.vert
0.210† −0.101 −0.00214
119 123 1.02
0.22 −0.12 −0.54
n.s. n.s. 3.4 × 10−4
−0.0338 −0.226 −0.00124
141 141 0.991
−0.028 −0.20 −0.41
n.s. n.s. 0.0095
n.s. 0.0223 n.s.
0.0920 −0.154 −0.00164
130 131 1.00
0.082 −0.15 −0.47
n.s. n.s. 1.5 × 10−5
Tb.Th*.vert.50th Tb.Th*.horz.50th Tb.Th*.vert.95th Tb.Th*.horz.95th
0.0174 −0.253 1.19 0.524
122 121 163 184
0.018 −0.30 0.47 0.21
n.s. n.s. 0.0021 n.s.
−0.110 −0.321 0.384 0.242
134 132 238 229
−0.11 −0.30 0.098 0.07
n.s. n.s. n.s. n.s.
n.s. n.s. 0.0432 0.0453
−0.0439 −0.280 0.800 0.410
128 126 199 205
−0.044 −0.29 0.23 0.13
n.s. 0.010 0.038 n.s.
Trabecular thickness distribution The 50th percentile (i.e. the median) of the trabecular thickness distribution did not change significantly with age for vertical or horizontal trabeculae for either women or men (Figs. 4A & B). The 95th percentile for the vertical trabecular thickness distributions increased significantly with age for women (r = 0.47, p=0.0021), but not for men (Fig. 4C). Actually, the 90th to the 99th percentiles for the vertical trabecular thickness distributions all increased significantly with age for women, but not for men. Moreover, the slope of the age-percentile fit lines increased monotonically from 0.88 μm/year (for the 90th percentile) to 1.46 μm/year (for the 99th percentile) for women, indicating a more pronounced increase in trabecular thickness with age for the thickest (highest percentile) vertical trabeculae. In contrast, the corresponding percentiles for the horizontal thickness distributions did not change significantly with age for either women or men (the 95th percentiles are depicted in Fig. 4D). Fig. 5 shows representative examples of the vertical thickness distribution for a 22-year-old woman (A), a 96-year-old woman (B), a 22year-old man (C), and a 90-year-old man (D). These vertical trabecular thickness distributions illustrate how the 95th percentile increased with age for the women, but not for the men. At the same time the median was similar for all four individuals, thus illustrating how the vertical thickness distribution changed shape with age for women, while remained almost unchanged with age for men. Discussion We here present a method for segmenting a set of 3D bone data into either vertical or horizontal voxels depending on the orientation of the bone in the local neighborhood and applied it to 79 human vertebral body bone specimens. In the present study we showed that the average trabecular bone volume fraction decreased significantly with age for both women and men, and that this age-related loss of bone density did not differ significantly between the two sexes. This is consistent with our previous findings for a subpopulation using 2D histomorphometry [26]. However, when the bone voxels were divided into horizontal and vertical voxels, we found that vertical bone was lost significantly faster with age for women than for men, whereas no sex-related differences were seen
for horizontal bone. In addition, the absolute loss of bone density was significantly larger for vertical oriented bone than for horizontal oriented bone for both sexes. In contrast, the relative loss of bone volume was similar for horizontal and vertical bone volume for women, whereas relatively more horizontal than vertical oriented bone volume was lost with age in men. The difference between the absolute and the relative bone losses is not surprising as the amount of vertical oriented bone volume was more than twice that of horizontal oriented bone volume. The present study showed that the average trabecular thickness determined with the direct method (i.e. assumption independent) Tb.Th* did not change significantly with age for either women or men. This is in accordance with the findings of Stauber and Müller, who investigated vertebral bone from 29 women and 38 men using μCT [8], and Liu et al. who investigated vertebral bone from 32 men using SEM [41]. However, Liu et al. did find a significant decrease in trabecular thickness from 20–29 years old men to 30–39 years old men, whereas no further decrease in trabecular thickness in older men was found. Furthermore, in the present study, we found that trabecular thickness determined using the parallel plate model Tb.Th decreased significantly with age for women and men when pooled. This is consistent with our previous finding using 2D histomorphometry, where we also found that Tb.Th decreased significantly with age when women and men were pooled [26]. The reason for this discrepancy between the findings for Tb.Th* and Tb.Th is most likely that the parallel plate model determines Tb.Th as 2 divided by the surface volume density, SV. With age the trabeculae are gradually converted from plate-like to rod-like (i.e. SMI increases with age), which will lead to a higher trabecular surface area per bone volume, i.e. a higher SV can be achieved with an unaltered trabecular thickness. As a consequence the trabecular thickness determined with the parallel plate model will increasingly be underestimated with age [42]. Although neither horizontal nor vertical trabecular thickness changed significantly with age, the horizontal to vertical trabecular thickness ratio decreased significantly with age for both sexes indicating that relatively more horizontal trabecular thickness is lost with age. However, the absolute trabecular thickness decreased significantly faster with age for horizontal trabeculae than for vertical trabeculae for women, whereas the loss of absolute trabecular thickness with age did not differ between horizontal and vertical trabeculae for men. It is interesting to note that the more rapid loss of vertical bone density in women than in men is not reflected in a similar more rapid loss of
52
J.S. Thomsen et al. / Bone 57 (2013) 47–55
12
225
A
D
200
10
Tb.Th*.vert (µm)
BV.vert/TV (%)
175 8 6 4
150 125 100 75 50
2 25 0
0 10 20 30 40 50 60 70 80 90 100
10 20 30 40 50 60 70 80 90 100
Age (years)
Age (years)
12
225
B
E
200
Tb.Th*.horz (µm)
BV.horz/TV (%)
10 8 6 4
175 150 125 100 75 50
2 25 0
0 10 20 30 40 50 60 70 80 90 100
10 20 30 40 50 60 70 80 90 100
Age (years)
Age (years)
1.0
1.2
BV.horz/BV.vert
F Tb.Th*.horz/Tb.Th*.vert
C 0.8
0.6
0.4
0.2
0.0
1.0
0.7
0.5
0.2
0.0 10 20 30 40 50 60 70 80 90 100
10 20 30 40 50 60 70 80 90 100
Age (years)
Age (years)
Fig. 3. Age-related changes of horizontal and vertical trabeculae. (A) Vertical trabecular bone volume. (B) Horizontal trabecular bone volume. (C) Ratio between horizontal and vertical bone volume. (D) Vertical trabecular thickness. (E) Horizontal trabecular thickness. (F) Ratio between horizontal and vertical trabecular thickness. Women are shown with open circles and dashed fit-lines, while males are shown with filled circles and solid fit-lines. Regression coefficients, correlation coefficients, and p-values for the linear relationships are shown in Table 1.
trabecular thickness in women than in men. The reason for this apparent paradox can be found in the trabecular bone structure. Either more vertical trabeculae are lost with age in women than in men, or more vertical plate-like trabeculae are converted into rod-like trabeculae in women than in men in order to account for the more pronounced loss of bone density in women than in men. The latter explanation is consistent with our findings for SMI, which showed that the conversion of plate-like trabeculae into rod-like trabeculae occurs significantly faster with age in women than in men. Macdonald et al. suggested that trabecular thinning might be a more common mechanism of bone loss in men owing to decreased rates of bone formation, whereas loss of trabecular elements might be more common in women owing to menopause-associated increases in osteoclastic activity leading to removal of entire trabeculae [43]. In the present study the loss of trabecular thickness with age was more pronounced in men than in women, although this was not statistically significant. Likewise, the loss of trabecular elements as assessed by Tb.N* was more pronounced in
women than in men, although this was borderline significant only. However, it should be born in mind that both SMI and Tb.N* were determined for the entire trabecular network and not for vertical trabeculae only. Further analysis of the trabecular thickness distributions showed that the thickest 5% of the vertical trabeculae (the 95th percentile) became significantly thicker with age for women, but not for men. Moreover, the age-related shift in the vertical trabecular thickness distribution towards thicker trabeculae in women was even more pronounced for the highest percentiles. At the same time the median and the mean trabecular thickness of both horizontal and vertical trabeculae was independent of age for both sexes. This indicates that the thickest vertical trabeculae underwent a compensatory thickening with age in women, but not in men. As mentioned above, we found that Tb.N* decreased borderline significantly faster in women than in men, indicating that more trabecular element may be lost with age in women than in men. This could explain why the compensatory hypertrophy is seen in
J.S. Thomsen et al. / Bone 57 (2013) 47–55
700
A
C
600
250
Tb.Th*.vert.95th (µm)
Tb.Th*.vert.50th (µm)
300
53
200 150 100 50
500 400 300 200 100
0
0 10 20 30 40 50 60 70 80 90 100
10 20 30 40 50 60 70 80 90 100
Age (years)
Age (years) 700
B
D
600
250
Tb.Th*.horz.95th (µm)
Tb.Th*.horz.50th (µm)
300
200 150 100 50
500 400 300 200 100
0
0 10 20 30 40 50 60 70 80 90 100
10 20 30 40 50 60 70 80 90 100
Age (years)
Age (years)
Fig. 4. Age related changes in 50th and 95th percentile for the horizontal and vertical trabecular thickness distribution. (A) 50th percentile (median) for the vertical trabeculae. (B) 50th percentile (median) for the horizontal trabeculae. (C) 95th percentile for the vertical trabeculae. (D) 95th percentile for the horizontal trabeculae. Women are shown with open circles and dashed fit-lines, while males are shown with filled circles and solid fit-lines. Regression coefficients, correlation coefficients and p-values for the linear relationships are shown in Table 1.
20
20
A
C 15
Frequency (%)
Frequency (%)
15
10
50th 95th
5
10
50th
5
0
95th
0 0
100
200
300
400
500
600
0
100
Tb.Th*.vert (µm)
200
300
400
500
20
20
B
D 15
Frequency (%)
15
Frequency (%)
600
Tb.Th*.vert (µm)
10
50th
5
95th
10
50th
5
0
95th
0 0
100
200
300
400
Tb.Th*.vert (µm)
500
600
0
100
200
300
400
500
600
Tb.Th*.vert (µm)
Fig. 5. Frequency distribution of Tb.Th*.vert for (A) a 22-year-old woman, (B) a 96-year-old woman, (C) a 22-year-old man, and (D) a 90-year-old man. The 50th percentile (median) is indicated with a dashed line and the 95th percentile is indicated with a solid line.
54
J.S. Thomsen et al. / Bone 57 (2013) 47–55
women but not in men, as fewer trabeculae are responsible for carrying the load in women with age leading to a higher load on the individual vertical trabeculae. Frost invited other researchers to provide proof that vertical trabeculae can strengthen and ‘thicken’ in adults under increased mechanical loading [18]. The present study provides an initial proof for this, as we demonstrated that the thickest vertical trabeculae became thicker with age, however, in women only. Nevertheless, without longitudinal data it is difficult to establish with certainty whether the vertical trabeculae actually thicken or there is a disproportional removal of thinner trabeculae resulting in the observed changes in the vertical thickness distribution. Whether these morphological differences between women and men influence fracture risk is not clear and merit further investigations. Bergot et al. analyzed vertebral bone from 31 women and 30 men at various ages and determined the mean trabecular thickness of both horizontal and vertical trabeculae using 2D granulometry [21]. They found trabecular thickness distributions similar to those observed in the present study, although the peak in the thickness distribution plot was higher than that observed in the present study probably due to a larger bin width. In contrast to the present study, they observed a slight shift to the left in the vertical thickness distribution histograms with age for both women and men. This discrepancy may be due to that the study by Bergot et al. was based on 2D methods, whereas the present study employed 3D methods. The method described in the present study is an extension of our previously introduced method for analyzing 2D histological sections [24]. A similar approach was taken by Mc Donnell et al., who investigated three vertebral bodies in 3D at a spatial resolution of 30–36 μm [13]. However, in the present work we found that for high resolution scans (in this case a spatial resolution of 18.5 μm) the original algorithm identifies tiny protrusions on the trabeculae as trabeculae of the orthogonal direction. We showed that these erroneously identified voxels could be reclassified without compromising the overall identification of the trabeculae by a simple median filtration. The presented method is voxel based i.e. the trabecular network is subdivided into horizontal and vertical oriented bone tissue rather than horizontal and vertical oriented trabeculae, although in most cases the outcome would be similar. Stauber and Müller [8,44–46] and Liu et al. [47–49] have both presented a method for subdividing the trabecular network into individual trabeculae, thus making it possible to classify each trabecula as a plate or a rod and to quantify its spatial orientation. Consequently, it would be interesting to compare the capability of our relatively simple voxel based method to that of the more complex trabecular based method of Liu et al. and Stauber and Müller in a future study. In addition, with slight modifications of the scanner software other measures such as e.g. SMI can also be determined for horizontal and vertical trabeculae separately using the method described in the present paper. We find that the presented method is easy to implement and can, at least for Scanco μCT scanners, be applied retrospectively, and – based on visual inspection – provides convincing results. A limitation of the present study is that we investigated an approximately 9-mm-thick bone slice from the central part of the vertebrae. We have previously shown that the trabecular bone volume and micro architecture is not homogeneously distributed in the vertical direction in human vertebral bodies [50]. Likewise, some studies have indicated that there are differences in the distribution of bone density and micro structure between the anterior and posterior regions of the vertebra [51–53]. In the present study we used a fixed global threshold filtration procedure for segmentation of the scans in accordance with the recommendations given in the current μCT guidelines [54]. Some researchers argue that application of more complex segmentation methods can be advantageous [55–57]. However, since no differences in the agerelated changes between women and men in trabecular bone ashdensity were found in the adjacent L3 [25], there is no reason to believe that the applied segmentation technique influenced the sex-related differences found in the present study. Finally, the partial volume effect,
which is an inherent limitation to the CT technique, could potentially bias the measurements although this limitation is most pronounced for voxels sizes larger than that used in the present study [54]. Conclusion In the present study we have described an easy method for segmenting a trabecular network into horizontal and vertical oriented bone. Using this method on 79 human lumbar bone samples, we have shown that vertical and horizontal bone decreased with age in both women and men, and that vertical oriented bone was lost more rapidly in women than in men. Furthermore, vertical as well as horizontal trabecular thickness were found to be independent of age, whereas the horizontal/vertical trabecular thickness ratio decreased significantly with age indicating a more pronounced thinning of horizontal trabeculae. Finally, we found indications that the age-related loss of trabecular elements resulted in compensatory hypertrophy of vertical trabeculae in women, but not in men. Disclosures All authors state that they have no conflicts of interest. Acknowledgments Andres Laib and Stephan Weiss, Scanco Medical AG, are gratefully acknowledged for help with IPL and decoding Scanco's AIM image file format. The μCT scanner was donated by the VELUX Foundation. The Helga and Peter Korning Foundation and the A.P. Møller Foundation for the Advancement of Medical Science are acknowledged for financial support of the study. Authors' roles: Conception and design: JST and AB. Acquisition of data: JST, ASN, and ENE. Analysis and interpretation of data: JST, ASN, ENE, AB. Drafting manuscript: JST. Revising manuscript: JST, ASN, ENE, and AB. Approved final version of manuscript: JST, ASN, ENE, AB. Responsibility for integrity of data analysis: JST. References [1] Ebbesen EN, Thomsen JS, Beck-Nielsen H, Nepper-Rasmussen HJ, Mosekilde Li. Lumbar vertebral body compressive strength evaluated by dual-energy X-ray absorptiometry, quantitative computed tomography, and ashing. Bone 1999;25:713–24. [2] Perilli E, Briggs AM, Kantor S, Codrington J, Wark JD, Parkinson IH, et al. Failure strength of human vertebrae: prediction using bone mineral density measured by DXA and bone volume by micro-CT. Bone 2012;50:1416–25. [3] Kleerekoper M, Villanueva AR, Stanciu J, Rao DS, Parfitt AM. The role of three-dimensional trabecular microstructure in the pathogenesis of vertebral compression fractures. Calcif Tissue Int 1985;37:594–7. [4] Dempster DW. Bone microarchitecture and strength. Osteoporos Int 2003;14(Suppl. 5):S54–6. [5] Räth C, Monetti R, Bauer J, Sidorenko I, Müller D, Matsuura M, et al. Strength through structure: visualization and local assessment of the trabecular bone structure. New J Phys 2008;10:125010. [6] Parkinson IH, Badiei A, Stauber M, Codrington J, Müller R, Fazzalari NL. Vertebral body bone strength: the contribution of individual trabecular element morphology. Osteoporos Int 2012;23:1957–65. [7] 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. [8] Stauber M, Müller R. Age-related changes in trabecular bone microstructures: global and local morphometry. Osteoporos Int 2006;17:616–26. [9] Bell GH, Dunbar O, Beck JS, Gibb A. Variations in strength of vertebrae with age and their relation to osteoporosis. Calcif Tissue Res 1967;1:75–86. [10] Jensen KS, Li Mosekilde, Le Mosekilde. A model of vertebral trabecular bone architecture and its mechanical properties. Bone 1990;11:417–23. [11] Li Mosekilde. Sex differences in age-related loss of vertebral trabecular bone mass and structure—biomechanical consequences. Bone 1989;10:425–32. [12] Atkinson PJ. Variation in trabecular structure of vertebrae with age. Calcif Tissue Res 1967;1:24–32. [13] Mc Donnell P, Harrison N, Liebschner MA, Mc Hugh PE. Simulation of vertebral trabecular bone loss using voxel finite element analysis. J Biomech 2009;42:2789–96. [14] van der Linden JC, Homminga J, Verhaar JA, Weinans H. Mechanical consequences of bone loss in cancellous bone. J Bone Miner Res 2001;16:457–65.
J.S. Thomsen et al. / Bone 57 (2013) 47–55 [15] Fields AJ, Lee GL, Liu XS, Jekir MG, Guo XE, Keaveny TM. Influence of vertical trabeculae on the compressive strength of the human vertebra. J Bone Miner Res 2011;26: 263–9. [16] Homminga J, Van-Rietbergen B, Lochmüller 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. [17] Fields AJ, Nawathe S, Eswaran SK, Jekir MG, Adams MF, Papadopoulos P, et al. Vertebral fragility and structural redundancy. J Bone Miner Res 2012;27:2152–8. [18] Frost HM. On the trabecular “thickness”-number problem. J Bone Miner Res 1999;14:1816–21. [19] Casuccio C. An introduction to the study of osteoporosis (biochemical and biophysical research in bone ageing). Proc R Soc Med 1962;55:663–8. [20] Henschke F, Pesch HJ. Spongiosastruktur von lendenwirbelkörpern und schenkelhals vergleichende analyse des altersabhängigen umbaues. Microsc Acta Suppl 1980;4:124–8. [21] Bergot C, Laval-Jeantet AM, Prêteux F, Meunier A. Measurement of anisotropic vertebral trabecular bone loss during aging by quantitative image analysis. Calcif Tissue Int 1988;43:143–9. [22] Jayasinghe JAP, Jones SJ, Boyde A. Three-dimensional photographic study of cancellous bone in human fourth lumbar vertebral bodies. Anat Embryol (Berl) 1994;189: 259–74. [23] Parfitt AM, Mathews CHE, 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 cellular mechanisms of bone loss. J Clin Invest 1983;72:1396–409. [24] 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. [25] Ebbesen EN, Thomsen JS, Beck-Nielsen H, Nepper-Rasmussen HJ, Mosekilde Li. Ageand gender-related differences in vertebral bone mass, density, and strength. J Bone Miner Res 1999;14:1394–403. [26] Thomsen JS, Ebbesen EN, Mosekilde Li. Static histomorphometry of human iliac crest and vertebral trabecular bone: a comparative study. Bone 2002;30:267–74. [27] Ebbesen EN, Thomsen JS, Beck-Nielsen H, Nepper-Rasmussen HJ, Mosekilde Li. Vertebral bone density evaluated by dual-energy X-ray absorptiometry and quantitative computed tomography in vitro. Bone 1998;23:283–90. [28] Thomsen JS, Ebbesen EN, Mosekilde Li. A new method of comprehensive static histomorphometry applied on human lumbar vertebral cancellous bone. Bone 2000;27:129–38. [29] Sikjaer T, Rejnmark L, Thomsen JS, Tietze A, Brüel A, Andersen G, et al. Changes in 3-dimensional bone structure indices in hypoparathyroid patients treated with PTH(1–84): a randomized controlled study. J Bone Miner Res 2012;27:781–8. [30] Hildebrand T, Laib A, Müller R, Dequeker J, Rügsegger P. Direct three-dimensional morphometric analysis of human cancellous bone: microstructural data from spine, femur, iliac crest, and calcaneus. J Bone Miner Res 1999;14:1167–74. [31] Hildebrand T, Rügsegger P. A new method for the model-independent assessment of thickness in three-dimensional images. J Microsc 1997;185:67–75. [32] Laib A, Rüegsegger P. Comparison of structure extraction methods for in vivo trabecular bone measurements. Comput Med Imaging Graph 1999;23:69–74. [33] Odgaard A, Gundersen HJG. Quantification of connectivity in cancellous bone, with special emphasis on 3-D reconstructions. Bone 1993;14:173–82. [34] Hildebrand T, Rügsegger P. Quantification of bone microarchitecture with the structure model index. Comput Methods Biomech Biomed Engin 1997;1:15–23. [35] Thomsen JS, Laib A, Koller B, Prohaska S, Mosekilde Li, Gowin W. Stereological measures of trabecular bone structure: comparison of 3D micro computed tomography with 2D histological sections in human proximal tibial bone biopsies. J Microsc 2005;218:171–9.
55
[36] Laib A, Barou O, Vico L, Lafage-Proust MH, Alexandre C, Rügsegger P. 3D microcomputed tomography of trabecular and cortical bone architecture with application to a rat model of immobilisation osteoporosis. Med Biol Eng Comput 2000;38:326–32. [37] Bresenham JE. Algorithm for computer control of a digital plotter. IBM Syst J 1965;4: 25–30. [38] Thomsen JS, Ebbesen EN, Mosekilde Li. Relationships between static histomorphometry and bone strength measurements in human iliac crest bone biopsies. Bone 1998;22:153–63. [39] Diem K, Lentner C. Scientific tables. 7th ed. Basle, J.R: Geigy S.A; 1970 . [40] Press WH, Teukolsky SA, Vetterling WT, Flannery BP. Numerical recipes in C — the art of scientific computing. Cambridge: Cambridge University Press; 1992 . [41] Liu GM, Xu CJ, Kong N, Zhu XM, Zhang XY, Yao Y. Age-related differences in microstructure, density and biomechanics of vertebral cancellous bone of Chinese males. Aging Male 2012;15:233–9. [42] Day JS, Ding M, Odgaard A, Sumner DR, Hvid I, Weinans H. Parallel plate model for trabecular bone exhibits volume fraction-dependent bias. Bone 2000;27:715–20. [43] Macdonald HM, Nishiyama KK, Kang J, Hanley DA, Boyd SK. Age-related patterns of trabecular and cortical bone loss differ between sexes and skeletal sites: a population-based HR-pQCT study. J Bone Miner Res 2011;26:50–62. [44] Stauber M, Müller R. Volumetric spatial decomposition of trabecular bone into rods and plates—a new method for local bone morphometry. Bone 2006;38:475–84. [45] Stauber M, Rapillard L, van Lenthe GH, Zysset P, Müller R. Importance of individual rods and plates in the assessment of bone quality and their contribution to bone stiffness. J Bone Miner Res 2006;21:586–95. [46] Stauber M, Müller R. A sensitivity analysis of the volumetric spatial decomposition algorithm. Comput Methods Biomech Biomed Engin 2007;10:25–37. [47] Liu XS, Sajda P, Saha PK, Wehrli FW, Guo XE. Quantification of the roles of trabecular microarchitecture and trabecular type in determining the elastic modulus of human trabecular bone. J Bone Miner Res 2006;21:1608–17. [48] Liu XS, Sajda P, Saha PK, Wehrli FW, Bevill G, Keaveny TM, et al. Complete volumetric decomposition of individual trabecular plates and rods and its morphological correlations with anisotropic elastic moduli in human trabecular bone. J Bone Miner Res 2008;23:223–35. [49] Liu XS, Zhang XH, Guo XE. Contributions of trabecular rods of various orientations in determining the elastic properties of human vertebral trabecular bone. Bone 2009;45:158–63. [50] Thomsen JS, Ebbesen EN, Mosekilde Li. Zone-dependent changes in human vertebral trabecular bone: clinical implications. Bone 2002;30:664–9. [51] Sandor T, Felsenberg D, Kalender WA, Brown E. Global and regional variations in the spinal trabecular bone: single and dual energy examinations. J Clin Endocrinol Metab 1991;72:1157–68. [52] 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. [53] Hussein AI, Morgan EF. The effect of intravertebral heterogeneity in microstructure on vertebral strength and failure patterns. Osteoporos Int 2013;24:979–89. [54] Bouxsein ML, Boyd SK, Christiansen BA, Guldberg RE, Jepsen KJ, Müller R. Guidelines for assessment of bone microstructure in rodents using micro-computed tomography. J Bone Miner Res 2010;25:1468–86. [55] Scherf H, Tilgner R. A new high-resolution computed tomography (CT) segmentation method for trabecular bone architectural analysis. Am J Phys Anthropol 2009;140:39–51. [56] Burghardt AJ, Kazakia GJ, Majumdar S. A local adaptive threshold strategy for high resolution peripheral quantitative computed tomography of trabecular bone. Ann Biomed Eng 2007;35:1678–86. [57] Dufresne T. Segmentation techniques for analysis of bone by three-dimensional computed tomographic imaging. Technol Health Care 1998;6:351–9.