Spontaneous vertebral fractures are a common skeletal pathology in humans. Seven hundred thousand cases are reported in the United States each year, and 15% of women and 5% of men will experience at least one spontaneous fracture of the vertebral body in their lifetime (Melton et al.,1992; Riggs and Melton,1995; Davies et al.,1996). Spontaneous vertebral fractures are most common in the midthoracic vertebrae, specifically in the T8 vertebra (Cooper and Melton,1996). The occurrence of spontaneous vertebral fractures in humans has been related to bone loss and osteoporosis (Lentle et al.,2007). Although osteoporosis is rare in nonhuman apes, there have been reported cases of wild chimpanzees with low bone mineral density in their femoral necks and lumbar vertebrae (Sumner et al.,1989; Gunji et al.,2003). Instances of spontaneous vertebral fractures, however, have not been reported in nonhuman hominoids (Lovell,1990; Gunji et al.,2003).
A fracture occurs when the loads applied to a bone exceed its strength. Spontaneous vertebral fractures often develop unrelated to a traumatic event (Homminga et al.,2004), suggesting that they are the result of a reduction in bone strength as opposed to overloading. Whole bone strength is determined by tissue material properties and morphology (Hernandez and Keaveny,2006). Bone tissue material properties are determined by molecular level characteristics such as mineralization and collagen composition. Bone morphology encompasses size, shape, and internal architecture. Tissue material properties have little variation among closely related species (Currey,2002); therefore, differences in whole bone strength between humans and the other hominoids will be largely due to morphology. For example, the vertebral body carries the majority of the load applied to vertebrae, and it has been suggested that the vertebral bodies in humans increase in size within the spine from superior to inferior due to the increase in load in more caudal regions of the spine (White and Panjabi,1990). However, a recent structural analysis of hominoid thoracic vertebrae suggested that while human vertebrae may be larger and more porous (lower bone volume fraction) than thoracic vertebrae of similarly sized apes, they have similar whole bone strength (Hernandez et al.,2009). The study by Hernandez and colleagues calculated differences in trabecular bone density but did not examine trabecular microarchitecture. Trabecular bone occupies the majority of the vertebral body and, in humans, carries ∼55–85% of the total load applied (Eswaran et al.,2006). Differences in trabecular bone microarchitecture among hominoids may influence vertebral whole bone strength and susceptibility of spontaneous vertebral fractures.
The strength and stiffness (resistance to fracture) of trabecular bone are highly correlated with bone volume fraction (bone volume/total volume) (Keaveny,2001); however, bone volume fraction does not account fully for trabecular bone strength (Rice et al.,1988; Hildebrand et al.,1999; Ulrich et al.,1999). Other measures of trabecular microarchitecture that are independent of bone volume fraction, such as degree of anisotropy (preferential alignment of the trabeculae) may influence bone strength and susceptibility to fracture. Differences in degree of anisotropy at nonvertebral fracture sites, such as the femoral neck, have been implicated in dissimilarity of fracture occurrence in individuals with similar bone volume fraction (Ciarelli et al.,2000). Analyses of human osteoporotic vertebral trabecular microarchitecture have shown that trabecular microarchitecture plays an important role in resistance to vertebral fracture (Bell et al.,1967; Kleerekoper et al.,1985; Homminga et al.,2004).
With the advent of microcomputed tomography (Feldkamp et al.,1989), it became possible to resolve the gross morphology of the trabecular bone structure at a very high resolution. Additionally, it allows for a three-dimensional view of what had been formerly analyzed in two dimensions (Fajardo and Muller,2001). Trabecular organization reflects the primary loading axes during bone growth (Wolff,1892; Martin et al.,1998), and microcomputed tomography studies of the proximal femur (Ryan and Ketcham,2002a,b,2005; Ryan and van Rietbergen,2005; Ryan and Krovitz,2006; Fajardo et al.,2007), calcaneus (Maga et al.,2006), and vertebrae (Agarwal et al.,2004; Fajardo et al.,2005) have shed light on the relationship of trabecular microarchitecture with locomotion and lifestyles of humans and nonhuman primates. The extant hominoids are diverse in their body size and anatomy, habitats and life history, and habitual locomotor patterns. These differences in activity and growth are presumably reflected in trabecular microarchitecture throughout the skeleton including the spine. Thus, any differences in trabecular microarchitecture among hominoids are likely to be related to differences in growth patterns and habitual loading among these species.
Currently, it is not known how trabecular microarchitecture in the thoracic spine differs among hominoids. It is possible that differences in trabecular microarchitecture between humans and nonhuman apes contribute to the difference in susceptibility to spontaneous vertebral fractures. The overall goal of this research is to determine how trabecular microarchitecture relates to normal form and function of the skeleton in Hominoidea; furthermore, we seek to determine how these relationships influence the incidence of skeletal pathology among this superfamily. In this study, we determine whether there is a difference in the trabecular microarchitecture of the vertebrae between humans and nonhuman hominoids.
MATERIALS AND METHODS
The eighth thoracic (T8) vertebral bodies of five hominoid (gibbon, orangutan, chimpanzee, gorilla, human) species (N = 6 per species, mixed sex) were studied (Table 1). Human vertebral specimens were taken from individuals who died of sudden illnesses or traumatic accidents, and nonhuman vertebral specimens were taken from wild-shot individuals. The human specimens ranged in age from 20 to 32 years old, and, based on epiphyseal fusion and tooth wear, all nonhuman specimens were young adults. Five orangutan specimens were from the Mammal Collection at the National Museum of Natural History, Washington, DC, and all remaining hominoid specimens were from the Hamann-Todd Osteological Collection at the Cleveland Museum of Natural History, Cleveland, Ohio. Because of the paucity of wild-shot orangutan specimens, two Pongo abelii and four Pongo pygmaeus were grouped together to represent the orangutan sample. Body mass for all specimens was estimated using femoral head dimensions (Ruff,1988; McHenry,1992; Hernandez et al., In Press). Estimated body masses of the individuals ranged from 6.75 to 201.40 kg.
Table 1. Study sample details
Estimated body weight (kg)
VOI volume (mm3)
Image Acquisition Using Microcomputed Tomography
The specimens were scanned at the Cleveland Clinic Foundation (Cleveland, Ohio) in a GE Locus eXplore RS microcomputed tomography machine (GE Healthcare, Milwaukee, WI). The Hylobates were scanned at a resolution of 26 μm, and the remaining specimens were scanned at 46 μm to create 16-bit, three dimensional images with 20 and 46 μm equant cubic voxels, respectively. The 26 μm images had a field of view of 40 mm diameter, and the 46 μm images had a field of view of 80 mm. These fields of view were sufficient to image the entire vertebral body of each specimen. The software package Microview Analysis+ 2.2 (GE Healthcare, Milwaukee, WI) was used to orient the vertebral body images into anatomical position and to choose the volume of interest (see below). The gibbon images were coarsened with Microview Analysis+ 2.2 to a 45 μm resolution for consistent image analysis with the other species.
Volume of Interest Selection
A hexahedral volume of interest (VOI) was placed in the anterior region of the vertebral body for analysis (Fig. 1). The area and shape of the VOIs in this study attempted to mimic a core of trabecular bone that would be used in a biomechanical test. Because the species included here range in size from very small (Hylobates) to very large (Gorilla), it was necessary to scale the VOI to ensure a similar region of the vertebral body was assessed. Choosing a parameter to scale against can be problematic, and it is necessary to choose one that is anatomically and biomechanically relevant (Fajardo and Muller,2001). Additionally, a sample of trabecular bone must be a minimum of 3–5 mm in smallest dimension (a continuum level representation) to ensure that the microarchitecture measures are characteristic of the trabecular bone as a material. A sample this size is also appropriate for direct biomechanical testing (Keaveny,2001).
The minimum width of the vertebral body in the transverse plane was chosen as the scaling parameter for the transverse and anteroposterior dimensions of the VOI (Fig. 2). Using the transverse mid-body plane, the VOI was placed as far anterior as possible without including the cortical shell (Fig. 1). Finally, the height of the VOI was taken as the full height of the vertebral body minus the cortical shell (Figs. 1 and 2).
A custom segmentation program was written using MATLAB (Mathworks, Natick, MA) to segment the images (separate bone from nonbone in a grayscale image). The images were first smoothed with a three dimensional Gaussian filter (7 × 7 × 7 voxel kernel, σ = 1 voxel) to account for partial volume effects and remove noise from the image (Kazakia et al.,2008). Each image threshold was determined manually and compared with the original, gray-scale image to ensure the most accurate delineation of the trabecular structure (Hara et al.,2002; Kazakia et al.,2008). Global thresholds determined in this manner have been shown to have little effect on microarchitectural measurements (Hara et al.,2002).
Trabecular Microarchitecture Analysis
The images were imported into Quant3D and 3DCalculator for analysis. Quant3D software (University of Texas-Austin) was used to determine bone volume fraction, trabecular thickness, trabecular separation, trabecular number, and mean intercept length degree of anisotropy (Ketcham and Ryan,2004). 3DCalculator software (Skyscan, Kontich, Belgium) was used to determine structure model index. Trabecular thickness and separation characterizes the average size of the trabeculae and spaces between the trabeculae, respectively, using a star volume distribution, and is the average thickness of the trabeculae (or space between the trabeculae) in the specimen per unit volume (Hildebrand and Ruegsegger,1997a). Trabecular number is the inverse of the mean distances between the mid-axes of the trabeculae (Hildebrand et al.,1999; Ketcham and Ryan,2004). Structure model index is the degree of rod-like versus plate-like trabeculae (Hildebrand and Ruegsegger,1997b) with higher values being more rod-like and lower values being more plate-like.
Degree of anisotropy is a characterization of the organization of the trabecular structure (Odgaard,1997). The value of degree of anisotropy indicates the degree to which the trabecular structure shows a preferential orientation—low values indicate less organization (isotropic), higher values indicate more organization (anisotropic). Although degree of anisotropy describes organization, it does not indicate in which direction the trabeculae are oriented; instead, fabric tensors are necessary to detect directionality (Odgaard,2001). Degree of anisotropy was calculated using the mean intercept length ellipsoidal model, and directionality was visualized using three dimensional star volume rose diagrams (Odgaard et al.,1997; Odgaard,2001; Ketcham and Ryan,2004).
Before comparative analyses, each microarchitecture measure was compared between females and males of each species using Mann-Whitney nonparametric tests and were determined to have no significant differences between the sexes. Accordingly, all parameters were sex-pooled for analysis. In addition, phylogenetic signal is unlikely to significantly influence regression analyses when comparing fewer than seven species, and all regression analyses were performed assuming independence among the species (Garland and Adolph,1994).
Differences among species were identified using one-way ANOVA tests, as well as nonparametric Kruskal-Wallis tests where appropriate (structure model index), with post hoc Holm multiple comparisons (Glantz,2005). Reduced major axis regressions of log transformed data were performed to determine the allometric relationship between the trabecular microarchitecture measures and body mass (Sokal and Rohlf,1995). Additionally, because many trabecular microarchitecture measures included in this study have been found to be correlated with bone volume fraction (Swartz et al.,1998; Fajardo et al.,2005), ordinary least squares regression and analysis of covariance (ANCOVA) were used to determine whether there were differences in trabecular microarchitecture measures among species after accounting for bone volume fraction. Post hoc power analyses for all tests were performed using a power threshold of 0.8. All statistical tests were performed using MINITAB 15, and significance thresholds were set at α = 0.05.
Analysis of variance results indicated that with the exception of degree of anisotropy, significant differences were observed among species in all of the trabecular microarchitecture measures; however, no significant differences in bone volume fraction were observed among the species in post hoc pairwise multiple comparison tests (Tables 2 and 3). Nonparametric Kruskal-Wallis test indicated that there were no significant differences among species in structure model index. Of the trabecular microarchitecture measures displaying significant differences (thickness, number, and separation), no species was consistently different from the others (Table 3). Three dimensional star volume rose diagrams indicated that the primary trabecular orientation of all species was in the superoinferior direction.
Table 2. Descriptive statistics of trabecular microarchitecture measures
Trabecular microarchitecture measure
Kruskal-Wallis nonparametric P-value, please see Results section.
Table 3. Holm pairwise multiple comparison test resultsa
Trabecular microarchitecture measures that are significantly different between pairs of species are listed. Please refer to Table 2 for acronyms.
Species-pooled investigation of allometry of the trabecular microarchitecture measure indicated that trabecular thickness, trabecular number, and trabecular separation to be related to body mass (P < 0.05). The allometric relationship of these significant correlations revealed that trabecular thickness ∝ body mass0.21 ± 0.03, trabecular number ∝ body mass−0.21 ± 0.03 and trabecular separation ∝ body mass0.26 ± 0.04 (coefficient ± SE). Bone volume fraction, structure model index, and degree of anisotropy were independent of body mass across species (P ≫ 0.05). Hence, in larger animals, trabeculae were thicker and decreased in number, but did not show differences in porosity, shape (rod-like versus plate-like), or anisotropy.
Across all species, bone volume fraction was found to be significantly correlated with trabecular number, trabecular separation, and structure model index (P < 0.05) using ordinary least squares regression (Table 4). Trabecular thickness displayed a trend suggesting a positive relationship with bone volume fraction (P = 0.12). These results suggest that as the bone volume increases trabeculae become more numerous and more plate like. No differences in the relationship between bone volume fraction and trabecular microarchitecture measures were found among all species.
Table 4. Species-pooled bone volume fraction ordinary least squares regressiona
Degree of anisotropy was not correlated with bone volume fraction in the nonhuman apes. However, in humans, degree of anisotropy was strongly negatively correlated with bone volume fraction (r2 = 0.85, P < 0.05, Fig. 3). This suggests that as trabecular bone porosity increases the trabecular bone becomes more anisotropic in humans, whereas no such relationship is observed in the nonhuman hominoids. Post hoc power analysis indicated that if there was a relationship between degree of anisotropy and bone volume fraction in nonhuman apes, it is so small that it would require more than 1,000 specimens to detect.
The purpose of this study was to determine the differences in the trabecular microarchitecture of the T8 vertebra in five extant hominoids with the long term goal of determining structural differences between human and nonhuman hominoids that are associated with susceptibility to spontaneous vertebral fracture. None of the microarchitecture measures included in this study completely distinguished between humans and other hominoids. However, the relationship between bone volume fraction and degree of anisotropy of vertebral trabecular bone differed between human and apes. In humans, trabecular bone had an increased degree of anisotropy in more porous trabecular bone, whereas there was no relationship between these two parameters in nonhuman apes.
This study had a number of strengths that lend confidence to our conclusions. First, we examined vertebrae from a sample of wild-shot, and not captive, apes. The use of wild-shot apes gave us confidence that the subjects had experienced the full range of locomotor modes possible in their natural habitats during their lifetimes. We used a sample set of six individuals per species which is large for a comparative study of hominoids due the rarity of wild-shot specimens from some nonhuman species providing reliability in our observations. Despite the large size of our sample, some of the correlation analyses (trabecular separation with body mass and trabecular thickness with bone volume fraction) showed insignificant trends and a low statistical power (30–40%). Because a post hoc power analysis indicated that the necessary sample sizes would be 2–3 times larger than our current study, a more powerful analysis would be logistically difficult with current museum collections. Second, we were able to estimate body mass individually for each specimen instead of relying on species means. Third, we performed three dimensional imaging on trabecular microarchitecture at a high resolution rather than relying on destructive histological sections or other two dimensional imaging techniques.
There are limitations to this study that must be considered when interpreting our results. First, scaling a volume of interest to analyze can be problematic. Fajardo and Muller (2001) note that when using individuals that span a large range in body size there is a danger of “oversampling” the smaller species by including dissimilar areas of the bone. To mitigate this potential limitation, our VOIs were scaled according to a direct measurement of the vertebral body and were placed in anatomically equivalent sections of the vertebral body.
Although differences existed in trabecular microarchitecture measures between species, no one microarchitecture measure differentiated the humans completely from the other hominoids. Trabecular thickness and trabecular number showed a significant but small dependence on body mass. The observed relationship between trabecular thickness and body mass was consistent with two dimensional trabecular diameter data in the humerus and femur of nonvolant mammals (Swartz et al.,1998). Using ordinary least squares regression, Swartz and colleagues found a negative allometric relationship between trabecular thickness and body mass. In our analysis, both ordinary least squares regression and reduced major axis regression of trabecular thickness with body mass also scaled with a similar negative allometry. Conversely, Fajardo and colleagues also found a positive trend of trabecular thickness with body mass in the lumbar vertebral bodies of strepsirrhine primates using microcomputed tomography, but found that trabecular thickness followed geometric similarity (trabecular thickness ∝ body mass0.33) (Fajardo et al.,2005). The difference between our results and the results of Fajardo and colleagues may be a product of the difference in body mass range (42 g–20.5 kg strepsirrhines versus 6.75–201.40 kg hominoids), a difference imposed by location within the spine (lumbar versus thoracic) or a difference in loads placed on the spine by habitual locomotor patterns.
Our results indicated that there was no significant difference in bone volume fraction between the hominoid species. The values for bone volume fraction found here were consistent with those studies utilizing vertebral bone from young adult humans (Nazarian et al.,2008). Furthermore, as expected, bone volume fraction was significantly higher than what is seen in studies that analyzed vertebral bone from individuals over the age of 50 (Homminga et al.,2004; Hulme et al.,2007). Spontanteous vertebral fractures in humans are classically attributed to a loss of bone strength due to a loss of bone mass. In apes, osteoporosis is rare, but in individuals that have developed osteoporosis spontaneous vertebral fractures have not been observed (Gunji et al.,2003). As a result, differences in susceptibility to vertebral fractures between humans and other apes cannot be due to bone mass alone.
Bone volume fraction is important in determining susceptibility to fragility fracture in human vertebrae, but it does not fully account for the strength of trabecular bone (Rice et al.,1988; Ulrich et al.,1999). For example, Ulrich and colleagues (1999) found that bone volume fraction could explain 52% of the variation in elastic modulus in human cancellous bone, but when differences in degree of anisotropy were accounted for as much as 74% of the variance in elastic modulus could be explained. It is evident that an anisotropic trabecular bone structure arises during development, and reflects the primary loads placed upon the whole bone (Tanck et al.,2001; Ryan and Krovitz,2006). Primary loads can arise from postural or locomotor differences and will be reflected in the anisotropic structure of the trabecular bone (Ryan and Ketcham,2002a,2005). Oxnard and Yang (1981) used Fourier transforms to analyze the trabecular orientation of the fourth lumbar vertebra and found that humans had more orthogonally (primarily craniocaudal and dorsoventrally oriented) trabeculae than the great apes, and attributed this to differing dynamic locomotor patterns. Conversely, Robson Brown and colleagues (2002) performed the same analysis on the entire thoracolumbar spine and found that humans, chimpanzees, and Neanderthal fossils were very similar in their trabecular orientation despite differences in locomotion, and the slight differences they did find were attributed to static postural loads as opposed to dynamic locomotor loading. That we did not find significant differences in the degree of anisotropy among species suggests that the T8 vertebra of each species is experiencing similar primary loading patterns. Fabric tensors indicated that the trabeculae were primarily oriented in a supero-inferior orientation, and this suggests that the primary loads experienced by the vertebral body are compressive in nature. These primary compressive loads are most likely due to similar postures and anatomical relationships at the T8 level: orthograde postures, dorsally convex thoracic spines, and a rib cage that limits movements (Shapiro,1991).
Although bone volume fraction and degree of anisotropy did not differ among species, there was a unique relationship between the two parameters in humans that was not seen in the apes (Fig. 3). As bone volume fraction decreased in humans, degree of anisotropy increases, potentially contributing to susceptibility to vertebral fracture. In humans, osteoporotic vertebral bone has a low volume fraction and is highly anisotropic (Bell et al.,1967; Kleerekoper et al.,1985; Homminga et al.,2004). High degrees of anisotropy can allow trabecular bone to be stronger and stiffer in one direction (usually the primary loading direction) with less bone volume, but will make the trabecular bone relatively weaker when loaded in other directions (Martin,2003; Homminga et al.,2004). There is evidence that degree of anisotropy may also be influential in spontaneous vertebral fractures. Vertebral trabecular bone from patients with osteoporosis exhibits an increase in vertically oriented trabeculae (Kleerekoper et al.,1985), and biomechanical analyses of vertebral bodies from patients with osteoporosis found that the trabeculae were more oriented in the primary loading direction making the whole bone less resistant to off-axis loads (Homminga et al.,2004). Furthermore, in human vertebral bone, the combination of low bone volume fraction and high degree of anisotropy results in trabeculae that are longer and more columnar (fewer horizontal trabeculae are connecting the vertical trabeculae). Bell and colleagues (1967) posited that bone loss in humans created vertical columns of trabeculae that were more subject to bending and buckling due to the reduced number of horizontal trabeculae, resulting in vertebrae that are weaker and more subject to spontaneous fracture.
Why there would be different trends in bone volume fraction and degree of anisotropy in humans and nonhuman apes is not clear. Although changes in bone volume fraction and degree of anisotropy are most evident in the pathogenesis of osteoporosis in older humans, the current study examined young adult humans and apes, and our results suggest that this relationship is evident in the human vertebral body before the onset of osteoporosis. In the nonhuman apes, however, degree of anisotropy is independent of bone volume fraction. If this trend were to continue in apes during age-related bone loss, we would not expect changes in degree of anisotropy or any associated increase in fragility. Oxnard and Yang theorized that the structure of the vertebral trabecular network is primarily orthogonal (due to compressive loads) in nature, and that differences in the loads applied across species would be “lesser modifications upon a single major theme” (Oxnard and Yang,1981, p 161). Of the five extant hominoid genera, four different primary locomotor patterns are displayed: bipedalism (humans), brachiation (gibbons), careful-climbing quadrumanualism (orangutans), and knuckle-walking quadrupedalism (chimpanzees and gorillas) (Fleagle,1999). Furthermore, arboreality is seen to some extent in all extant hominoid species with the exception of humans (Fleagle,1999). Because of the nonuniform placement and resistances inherent in occupying three-dimensionally complex arboreal substrates, it is likely that the spines of climbing animals would be regularly subjected to loading in more varied directions during bouts of high energy travel than would terrestrially bipedal humans. The subtle differences in the loads induced by these locomotor patterns may be different enough to allow for the maintenance of transverse trabeculae in the nonhuman species as bone volume fraction decreases, thereby maintaining a more isotropic structure. Further investigation into the effects of aging on vertebral bone trabecular microarchitecture of apes would be necessary to support this hypothesis.
The authors would like to thank Dr. Yohannes Haile-Selassie and Lyman Jellema of the Cleveland Natural History Museum and Dr. James Mead, Dr. Richard Thorington and Linda Gordon of the Smithsonian Institution National Museum of Natural History for use of museum specimens, Dr. Amit Vasanji of the Cleveland Clinic Foundation for use and assistance with the microcomputed tomography scanning and analysis, CJ Slyfield for assistance with segmentation and image creation and John Schlueter for assistance with images.