SEARCH

SEARCH BY CITATION

Keywords:

  • bone adaptation;
  • subchondral bone;
  • density;
  • posture

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED

Spatial patterning in the apparent density of subchondral bone can be used to discriminate between species that differ in their joint loading conditions. This study provides an experimental test of two hypotheses that relate aspects of subchondral apparent density patterns to joint loading conditions. First, the region of maximum subchondral apparent density (RMD) will correspond to differences in joint posture at the time of peak locomotor loads; and second, differences in maximum density between individuals will correspond to differences in exercise level. These hypotheses were tested using three age-matched samples of juvenile sheep. Two groups of five sheep were exercised, at moderate walking speeds, twice daily for 45 days on a treadmill with either a 0% or 15% grade. The remaining sheep were not exercised. Sheep walking on the inclined treadmill used more flexed knee postures than those in the level walking group at the time of peak vertical ground reaction forces. Kinematic measurements of knee posture were compared with knee postures estimated from the spatial position of the RMD on the medial femoral condyle. Our results show that the difference in the position of the RMD between the incline and level walking groups corresponded to the difference in knee postures obtained kinematically; however, exercised and nonexercised sheep did not differ in the magnitude of apparent density. These results suggest that patterns of subchondral apparent density are good indicators of the experimental modifications in joint posture during locomotion and may, therefore, be used to investigate differences between species in habitual joint loading. Anat Rec, 291:293–302, 2008. © 2008 Wiley-Liss, Inc.

Assessing the relationships between some characteristic of bony morphology and locomotor behavior is a common goal among functional morphologists, particularly for those with an interest in the fossil record. Reconstructing habitual joint posture, in particular, can provide information about the roles that limbs play in supporting the body and generating propulsive forces (Biewener, 1983; Polk, 2002). Fortunately, many bone properties respond to differences in mechanical loading conditions during an animal's lifetime (see reviews in Pauwels, 1980; Lanyon and Rubin, 1984; Martin et al., 1998; Carter and Beaupre, 2001). While the theoretical basis for such changes has been known for over a century, current research focuses on which types of mechanical signals are best transduced by bone cells (see review in Zernicke and Judex, 1999), mechanical, genetic and hormonal effects on how this mechanotransduction takes place (e.g., Damien et al., 1998; Zaman et al., 2000, 2006; Hsieh and Turner, 2001; Whitfield, 2003) and how bone in different regions responds to differences in loading direction and/or magnitude.

For example, trabecular struts and plates have long been argued to respond to loading conditions (Wolff, 1892; Pauwels, 1980), and recent work with either finite element modeling or experimental studies have demonstrated correspondence between trabecular structures and load orientations in the human calcaneus and femur (Carter and Beaupre, 2001; Gefen and Seliktar, 2004), as well as in the zygomatic arch of pigs (Teng et al., 1997), the calcaneus of potoroos (Biewener et al., 1996), and the distal femur of guinea fowl (Pontzer et al., 2006). Bone curvature is also known to change ontogenetically and in response to applied muscular and gravitational loads (Lanyon, 1980; Richmond, 1998; Jungers et al., 2002; Main and Biewener, 2006), and bone cross-sectional properties are also well known to respond to loading conditions during ontogeny (Goodship et al., 1979; Lanyon et al., 1982; Carrier, 1983; Jaworski and Uhthoff, 1986; Haapasalo et al., 1994; Pearson and Lieberman, 2004), although the effects are dependent on many factors including age, hormonal levels, and initial bone status (Forwood and Burr, 1993; Damien et al., 1998; Lee et al., 2003; Lieberman et al., 2003; Ruff, 2003a, b; Devlin and Lieberman, 2007). Although this list is not comprehensive, these studies demonstrate that many aspects of bony morphology respond to bone loading conditions during an animal's lifetime. Assessing the adaptability of additional bone properties allows more comprehensive understanding of bone and joint function and established relationships are also useful for functional morphologists with interests in interpreting the behavior of extinct species. The current study is conducted with the goal of evaluating whether or not subchondral apparent density of the distal femur responds to differences in load orientation during a relatively short-duration experiment in an ovine model. If successful, these results will contribute to understanding the joint loading conditions in both extant and subfossil mammals.

Subchondral bone is the thin plate of bone that is situated between the articular cartilage and trabecular bone. Because of its position at this interface, it may serve to transmit and distribute loads to the trabeculae and may also have some load attenuating function (Radin et al., 1970a; Radin and Paul, 1971; Simon et al., 1972; Hoshino and Wallace, 1987). The responses of subchondral bone to joint loading are not completely understood, but it appears to respond in a manner similar to other bony tissues (Radin and Paul, 1971; Simon et al., 1972; Eckstein et al., 1995; Murray et al., 2001; Burr and Radin, 2003), and while bone mineral density is largely influenced by heredity (Krall and Dawson-Hughes, 1993), other environmental factors such as exercise play significant roles (e.g., Forwood and Burr, 1993; Magkos et al., 2007). Increasing bone density confers several advantages, such as increased strength in compression, compressive modulus, fatigue life, resistance to crack initiation, and tensile strength (Carter and Hayes, 1977; Wright and Hayes, 1977; Wall et al., 1979; Currey, 1988, 2002; Rice et al., 1988).

In any discussion of densities, it is important to clarify what quantities are being assessed and to distinguish between the true density for any given material (mass/volume) and the apparent density that is obtained through analysis of whole structures (e.g., whole bones including internal spaces) that contain several different types of tissue. In this study, we quantify apparent density of the subchondral bone in the distal femur.

Density variation in subchondral plates has been studied for a variety of clinical and comparative biological reasons. Changes in the density and stiffness of subchondral bone may be factors in the initiation of osteoarthritis. Initial work by Radin and colleagues suggested that, as a consequence of thickening and repair of trabecular microfractures, the subchondral plate was stiffened and stress increases within the overlying cartilage combined with repeated impact loading led to deterioration of the cartilage (Radin et al., 1972, 1973). Additional studies in humans and animal models have confirmed that there are density-related changes in the subchondral bone in individuals with osteoarthritis (Radin et al., 1970b; Radin and Rose, 1986; Brandt et al., 1991; Grynpas et al., 1991; Dedrick et al., 1993; Matsui et al., 1997; Hayami et al., 2004; Wang et al., 2005), but whether the subchondral changes precede or follow the osteoarthritis remains a matter of continuing debate (Burr and Radin, 2003; Wang et al., 2005).

Müller-Gerbl and colleagues introduced a procedure called computed tomography osteoabsorptiometry (CT-OAM), which has been widely used to characterize patterns in apparent density in the subchondral bone of several human joints, including the hip (Müller-Gerbl et al., 1992; Eckstein et al., 1994; Von Eisenhart-Rothe et al., 1999), patella (Eckstein et al., 1992, 1993; Müller-Gerbl et al., 1992), tibial plateau (Mockenhaupt and Koebke, 1988; Ahluwalia, 2000), coraco-acromial arch (Oizumi et al., 2003), scapular glenoid fossa (Müller-Gerbl et al., 1992), wrist (Giunta et al., 1999, 2004; Hoogenbergen et al., 2002), and metacarpophalangeal joints (Meirer et al., 2004). The majority of these studies have clinical applications, and their goals have primarily been to evaluate the patterns of normal and pathological load magnitudes. In other mammals, researchers have examined variation in subchondral density in the elbow joint of dogs (Samii et al., 2002), as well as the tibial plateau and wrist joint of primates (Ahluwalia, 2000; Carlson and Patel, 2006; Patel and Carlson, 2007), to characterize differences in load magnitude, load distribution, and posture. In addition, Müller-Gerbl et al. (1993) have observed increased in subchondral apparent density in athletes compared with ordinary people and elderly persons. This finding suggests that subchondral bone density may change with levels of exercise or loading magnitude.

Together, these studies demonstrate that analyses of subchondral bone provides information about the loading histories of joint surfaces (see also Fischer et al., 1995). What has been missing from these previous analyses, and what would make this technique most useful for broader comparative analyses of joint posture, is an experimental test to determine whether altering joint postures in a known manner will result in predictable changes in patterns of subchondral apparent density. This study will test the hypothesis that the position of maximum subchondral apparent density will correspond to the posture used when joint loads are highest (Fig. 1), and we attempt to determine whether exercise alters the magnitude of peak apparent density in the distal femur.

thumbnail image

Figure 1. Schematic illustrations of sagittal sections through sheep medial femoral condyles. A,B: Arrows represent the regions of the joint surface that will be loaded with habitual use of relatively flexed (A) and relatively extended (B) postures. The large arrow represents the center of contact between the tibia and femur, which should correspond to the location of the greatest joint force application.

Download figure to PowerPoint

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED

This experiment was conducted using an ovine model because sheep are tractable research subjects and are easily trained to locomote on treadmills. The knee joint was selected for analysis because the knee is one of the major limb joints and knee extensors are important for resisting gravity; the femoral condyles are highly convex, allowing loading on different parts of the joint surface with alterations in hindlimb posture; and because, alterations in substrate orientation are known to influence knee posture during stance (Vilensky et al., 1994; Smith and Carlson-Kuhta, 1995; Roberts et al., 1997; Jayne and Irschick, 1999; Stevens, 2003; Pontzer et al., 2006). The medial femoral condyle (MFC) was selected for analysis because it bears more weight than the lateral femoral condyle in other quadrupeds (Ahluwalia, 2000), and the orientation of the ground reaction force relative to the knee in sheep suggests that this should be similar in sheep as well (Polk, unpublished data). Thus, the medial condyle should record the effects of loading in a study of relatively short duration.

Fifteen juvenile domestic sheep (Ovis aries), housed at the University of Illinois Sheep and Beef Cattle facility, were divided into three groups: incline (n = 5), flat (n = 5) and control (n = 5). Subjects in the incline and flat groups were trained to walk on a motorized treadmill (Star Trac 4000HR, Star Trac, Irvine, CA) at moderate walking speeds (Table 1). Sheep were exercised twice daily (20 min/bout) for 45 days. Exercise bouts were separated by at least 6 hours to ensure effective recovery in cells responsible for transducing the mechanical stimuli (Robling et al., 2002). Control sheep were not exercised. When not being exercised, sheep were housed in a 6-m × 8-m pen. No restrictions were placed on their diet or water intake. While in the pen, the sheep were able to move freely. They would walk, run, and occasionally leap. However, their ability to engage in sustained (>30 s) locomotor activities was limited in this space, so the periods of treadmill exercise involved substantial increases in the frequency of limb loading, and the differences in exercise are likely to have had greater osteological effect than the activities in the pen. All animal subjects were treated humanely, and all procedures were approved by the UIUC Institutional Animal Care and Use Committee (protocol 04289).

Table 1. Number of strides and mean speed for each subjecta
SubjectnSpeed (SD)Mean (SD)
  • a

    Speed was controlled for the treadmill subjects and kinematically measured for the control sheep (see text for details).

Control m/secm/sec
127311.6(0.3)1.58(0.35)
132291.4(0.5) 
13952.0(0.6) 
14121.1(0.1) 
151171.8(0.5) 
Incline   
65151.211.21(0.25)
62151.34 
63151.56 
157200.98 
140200.98 
    
Flat   
67151.561.19(0.25)
64151.34 
133200.98 
143201.03 
128201.03 

Two-dimensional (2D) kinematic data were collected for each individual using a 30-Hz digital video camera (Sony USA) with 1,000-Hz shutter speed to reduce motion blur. Each subject was shorn and the skin overlying bony landmarks at six major limb joints was marked (anterior superior iliac spine, greater trochanter, lateral epicondyle of the femur, lateral malleolus, distal metatarsal, and anteroposterior center of hoof). Treadmill speeds used for collection of 2D kinematic data were identical to those used during exercise bouts. Kinematic data for control sheep were obtained for each sheep as it walked along an 8-m concrete walkway behind a Plexiglas barrier. Speed was not controlled, and subject cooperation varied, permitting between 2 and 31 independent strides to be obtained for each subject (Table 1). Speed for these subjects was variable and depended upon their motivation to move along the walkway. Kinematic measurement of joint posture was obtained when the hoof passed beneath the hip, that is, when the vertical ground reaction forces should reach a maximum (Biewener, 1983; Polk, 2001; Polk, unpublished data for different sample of sheep). Speed for control subjects was estimated by measuring the distance traveled by the greater trochanter through a complete stride (touchdown of one limb to subsequent touchdown of the same limb), as well as the time necessary to complete the stride. Relevant video frames were isolated using FinalCut Pro (Apple, Cupertino, CA), and images were digitized using tpsdig2 (Rohlf, 2006). Knee angle was measured as the angle between the greater trochanter, lateral epicondyle, and lateral malleolus markers. Following the collection of kinematic data, all subjects were euthanized humanely limbs were dissected, muscles were removed, and their limb bones were frozen.

Apparent density measurements were obtained using medical computed tomography (CT; General Electric Lightspeed Plus) at the Mount Auburn Hospital, Cambridge, MA, and were stored as 16-bit DICOM images. Specimens were completely thawed before the collection of CT data (thaw time was approximately 2 days), and all were scanned axially with the long axis of the bone perpendicular to the CT source/sensor plane to maximize the number of slices obtained per specimen. Pixel size on each slice was 0.234 mm × 0.234 mm, slice thickness was 0.625 mm. A soft-tissue CT reconstruction algorithm was used during the acquisition of CT data. This algorithm ensures that apparent density is related linearly to the intensity of the grey of the image with black indicating air-density and white indicating greater density. Apparent bone density is quantified in Hounsfield units (H, also called CT number), where apparent density values are expressed relative to water density (water density = 0H; Ruff and Leo, 1986). Air has an apparent density of −1000H and apparent densities for subchondral and compact bone range from approximately 150H to more than 2000H. The Mount Auburn CT system is calibrated daily to ensure the accuracy of density measurements. CT images were imported into AMIRA software for analyses of apparent density (v. 4.1, Mercury Computing Systems, Chelmsford, MA). In AMIRA, all CT slices were compiled to create a virtual 3D representation of the femur. This virtual 3D structure preserves the geometry of the whole bone and allows quantification of femoral shape and the position of the MFC relative to the shaft (see below). To quantify the position of the RMD on the MFC, a single, oblique, 2D slice was obtained through the long axis of the MFC for each specimen (slice thickness = 0.234 mm). To ensure consistency in slice orientation across individuals, the plane of this slice was oriented to pass through (i) the long axis of the MFC, and (ii) the medial lip of the patellar groove.

Sixteen-bit CT images record up to 216 levels of gray, and the human eye cannot readily determine the boundaries between regions of differing density without assistance. Thus, density variation on the subchondral surface was quantified by applying a color map to the scaled apparent density data in the DICOM images of the subchondral bone. AMIRA (v4.1) permits users to manually select the range of Hounsfield units to display (comparable to setting the CT level and window) and to divide this range into a maximum of 256 separate colors (bins). Maximum and minimum densities along the subchondral surface were assessed to the nearest 25 H. Densities less than or equal to the minimum found on the subchondral surface were assigned to bin 0, while densities greater than or equal to the maximum found on the subchondral surface were assigned to bin 255. The remaining density range (254 bins) was divided into eight regions of differing density (the highest and lowest density regions had one fewer bin than did the remaining six). Images of the oblique 2D slice were exported for digitizing in tpsdig2 (Rohlf, 2006). The highest density region is termed the RMD. The position of the RMD was measured on the MFC in two different ways (Fig. 2): as the angle between the posterior extent of the MFC, the point at the center of the line segment connecting the proximal and distal extents of the articular surface on the MFC and either (a) the anterior extent of the RMD (θ), or (b) the center of the RMD (μ). Both measurements were obtained to determine which one most closely matched the kinematic measurements. Coordinate measurements from tpsdig2 were exported, and angles were calculated using the law of cosines.

thumbnail image

Figure 2. This figure illustrates the methods for measuring the position of the region of maximum subchondral apparent density (RMD) on the medial femoral condyle. Whole bone geometry was used for three-dimensional (3D) measurement of the angle describing the orientation of the medial femoral condyle relative to the long axis of the bone (ρ = ∠DCB). The inset is a 2D slice through the medial femoral condyle with regions of differing density indicated by colors (highest density in red, lowest density in black). Angular measurements (μ, θ) describe the position of the center (μ = ∠BCF) and anterior extent (θ = ∠BCE) of the region of maximum density, respectively. Angle ρ was added to angles μ and θ to measure the angles of the center and anterior extents of the RMD relative to the long axis of the bone, respectively.

Download figure to PowerPoint

To determine the position of the RMD on the entire femur, and to estimate knee joint posture in a manner that is more comparable to kinematic measurements, it is necessary to quantify the location of the 2D slice on each femur. This was accomplished by combining the angle(s) describing the position of the RMD relative to a line connecting the proximal and distal ends of the MFC (from the 2D oblique slice), with an angle that describes the orientation of this line on the MFC relative to the long axis of the bone (in 3D). The 3D coordinate data were obtained from the CT data for each bone using a combination of AMIRA, to generate a dense surface mesh, and Landmark (Wiley et al., 2005) to place landmarks on this 3D surface. The 3D landmarks were placed on the superior margin of the proximal end of the greater trochanter (corresponding to the most posterodorsal projection of the greater trochanter when the femur is placed in anatomical position), and the proximal and distal extents of the MFC (Fig. 2). The midpoint of the MFC was calculated from these coordinates and this allowed calculation of the orientation of the medial femoral condyle relative to the long axis of the femur. Knee joint angles were estimated as the sum of the (a) orientation of the MFC relative to the long axis of the femur (Fig. 2, ρ) and (b) angular position of the RMD relative to the MFC (Fig. 2, μ or θ).

Because of the small sample size, differences between groups were assessed using nonparametric pairwise Mann-Whitney U-tests. All statistics were calculated in Systat (v.9, Systat Software Inc., San Jose, CA), with a type one error rate of α=0.05.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED

The central hypothesis of this research is that kinematically measured knee postures can be accurately reconstructed from the position of the RMD on the subchondral bone. Kinematic measurements of knee posture are shown in Figure 3A. As predicted, the sheep using the inclined treadmill exhibited significantly more flexed knee postures at the time of peak vertical force application than those using the flat treadmill (U = 25; P = 0.009). The difference between mean knee joint angle for these groups was 13.5°. Neither of the exercised groups differed significantly from the nonexercised control group (Control-Incline: U = 17; P = 0.347; Control-Flat: U = 5; P = 0.117). Variation around the mean was very low for the flat treadmill walkers and was substantially higher for the control subjects and incline walkers.

thumbnail image

Figure 3. A: Kinematic measurement of joint posture. B: Estimate of joint posture obtained by measuring the angle to the center of the region of maximal bone density (RMD; angle μ in Fig. 2). C: Estimate of joint posture obtained by measuring the angle to the anterior extent of the RMD (angle θ in Fig. 2). For B and C, smaller angles correspond to more posteriorly positioned RMD, which is consistent with more flexed limb postures during loading.

Download figure to PowerPoint

A similar pattern of differences is observed when comparing the positions of the subchondral RMD between groups of sheep (Fig. 3B,C). The center of the RMD is more posteriorly positioned for the sheep of the incline group, and more anteriorly positioned for the sheep that were exercised on the flat treadmill. This difference was statistically significant (U = 23; P =0.028), and corresponds to a difference in knee posture of 10.0°. Neither of the exercised groups of sheep differed significantly in knee posture from the nonexercised control group (Control-Incline: U = 18; P = 0.251; Control-Flat: U =7; P = 0.251). Using the anterior extent of the RMD, the flat and incline groups differed significantly (U = 22; P = 0.047), while the control group did not differ significantly from either the incline or flat groups (Control-Incline: U = 13; P = 0.917; Control-Flat: U = 19; P = 0.175).

The relative difference in inferred knee posture between incline and control groups is very similar to the difference obtained from kinematic measurements (difference of 3.5°). However, knee postures inferred from the center of the RMD substantially underestimate kinematically measured knee postures in this sample (Fig. 3A,B). For example, mean knee posture inferred from the center of the RMD for the incline group is 75.7°, while kinematically measured knee posture for this group is 120.2°. The anterior extent of the RMD performs better at estimating the kinematically measured knee posture, with the difference between incline group being approximately 16° (Fig. 3A,C).

The speeds used by the control subjects tended to be higher than in the flat or incline groups. The difference was significant for the control and flat groups (U = 22, P = 0.047), and approached but did not attain statistical significance for the control and incline groups (U = 21; P = 0.075). The flat and incline groups did not differ significantly (U = 13; P = 0.915).

Peak density values in the subchondral bone were predicted to be higher in both exercised groups than in the nonexercised control group. While the means of the exercised groups are higher than the control group (Fig. 4), no group differences in means were statistically significant (Control-Incline: U = 9.5; P = 0.530; Control-Flat: U = 7.5; P = 0.293; Flat-Incline: U = 15.5; P = 0.528).

thumbnail image

Figure 4. Differences in peak apparent density values between incline, flat, and control groups.

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED

The principal goal of this study was to experimentally test whether the patterns of subchondral apparent density could be used to infer differences in joint posture. Overall, there is good correspondence between certain density-based and kinematic results, suggesting that the selected characteristics can be used to infer differences in joint posture.

The difference in posture between the exercised groups was obtained by using treadmills that were either inclined or horizontal. The sheep exercised on inclined treadmills used more flexed knee postures when their hoof passed beneath their hip joint, compared with those that were exercised on horizontal treadmills. This result supports postural observations and kinematic measurements on other species using inclined and flat substrates for locomotion (Vilensky et al., 1994; Smith and Carlson-Kuhta, 1995; Roberts et al., 1997; Jayne and Irschick, 1999; Stevens, 2003; Pontzer et al., 2006).

The pattern of difference between groups was reflected in the position of maximal subchondral apparent density with the sheep exercised on the inclined treadmills showing a more posteriorly positioned RMD than the sheep exercised on the flat treadmill. This result was obtained when the position of the RMD was assessed by measuring its center as well as its anterior-most extent. The kinematically measured difference between incline and flat treadmill groups (13.5°) was replicated satisfactorily using both density-based methods, which gave consistent estimates (10° or 16° depending on the method). This difference (i.e., between kinematic- and density-based measurements) is small and well within the range of measurement error for any of the techniques (see below). While the difference between groups was consistent for each of the density-based methods, the absolute values for estimates of joint posture differed substantially. Both density-based techniques underestimated the kinematically measured joint angles, but the anterior extent of the RMD provided the more accurate reconstruction of knee joint angle. This discrepancy between density based estimates may be due to several factors. First, we had assumed that the center of the region of highest density might correspond to the center of pressure within the joint with pressure decreasing both anterior and posterior to this center (see Fig. 1). That the center of the RMD performed poorly at reconstructing joint posture suggests that this is not the case. Instead, because the anterior extent of the RMD provides more accurate reconstruction of the joint posture, we suggest that this anterior point may be closer to the center of pressure when pressures are highest. That a significant region of dense bone exists posterior to this point suggests that the sheep may be loading their bones in more flexed postures at other times.

In this study, we have tested a relatively simple hypothesis: that there is a relationship between the patterns of subchondral apparent density and the postures used during peak locomotor loading. We restricted our analysis to the time of peak force application to the limb, because this is likely to have greatest osteological effect. However, the patterns of density on the joint surface likely reflect additional variables (e.g., loading magnitudes, accumulated duration of load, and frequency). The substantial variation that exists in the position of the RMD in control sheep also attests to the fact that joint loading can be quite variable under normal circumstances. We did not make systematic observations of the locomotor behavior of the sheep while in the pen so we cannot comment further on the possible sources of this variation. Potential future analyses might analyze the range of motion during loading activities and determine whether this is correlated with variation in the position of the RMD.

The discrepancy between the kinematically measured joint angles and density-based methods is due primarily to the difference in measurement techniques. For example, the kinematic measurements were obtained using a marker on the distal tibia, which was obviously not used for the density-based methods. While there is only slight tibial curvature in sheep, this difference in methodology would result in consistent differences in joint angle from the density-based measurements (which cannot account for the location of the distal tibia). Correction for this effect has not been undertaken. Other factors, such as skin movement artifact may have larger effects and limit the accuracy with which true joint motion can be measured. This effect will vary by species, joint, and probably across individuals, but skin displacements of up to 20 mm have been reported for humans during several cyclic activities (e.g., Lundberg, 1996; Fuller et al., 1997; Reinschmidt et al., 1997). The magnitude of the induced error will certainly depend on the direction of this movement, but 20 mm error for a knee marker could cause discrepancies between measured and true posture of up to 20 degrees.

Previous work has suggested that differences in subchondral apparent density between subject populations might be explained by differences in exercise levels (e.g., Müller-Gerbl et al., 1993), and studies of the etiology of osteoarthritis have hypothesized a relationship between increased subchondral density and loading (e.g., Radin et al., 1970b; Radin and Rose, 1986), The relationship between exercise and subchondral apparent density was also assessed in this study. Ground reaction force magnitudes and midshaft bone strains vary in proportion to the speed of locomotion (Alexander and Jayes, 1980; Biewener and Taylor, 1986), so knee joint contact forces should respond in a similar way. However, in this study, the subjects exercised on the treadmills did not show increased peak densities compared with the nonexercised controls. This result was probably a consequence of the fact that the exercised sheep used only moderate walking speeds. Thus, knee joints of the exercised sheep may not have experienced higher forces than the control sheep. It should also be noted that, while the control sheep used faster speeds during kinematic measurements (Table 1), these speeds were higher than any of the sheep used during habitual locomotion in their enclosure, and as a consequence, we do not believe that the groups differed substantially in the magnitude of the loads that they experienced. Further studies should be conducted with animals exercised at significantly higher speeds than those used by the control subjects, and where quantification of the applied loads is possible.

We cannot claim that the exercised group differed from the control group in loading magnitude. However, we can state qualitatively, that the groups differed in the number of loading cycles that they experienced each day and in the frequency of the loading cycles they experienced during the exercise periods (i.e., the control sheep did not increase their level of activity while the exercised sheep did). The fact that peak densities did not differ between groups suggests either that the bone is not responding to changes in loading frequency or number of loading cycles, or alternatively that our levels of exercise were not sufficient to generate a significant osteological response.

The fact that density is modulated by joint loading may have implications for osteoarthritis research. Early models for the etiology of osteoarthritis have identified bone density changes as playing a causal role with increased subchondral density leading to increased subchondral stiffness and ultimately to the deterioration of articular cartilage (Radin et al., 1970b; Radin and Rose, 1986). However, recent work has questioned this relationship and suggested that increases in subchondral apparent density are really an artifact of increased subchondral thickness, and that, because subchondral stiffness is not increased (Grynpas et al., 1991; Burr and Radin, 2003), the increases in apparent density may be incidental to the development of arthroses (Burr and Radin, 2003). This debate continues. Our results show that the location of density increase is affected by joint posture, and our results also suggest that normal physiologic loads can produce changes in apparent density. It seems reasonable that such increases in apparent density would be an adaptive response to microdamage on the articular surface (c.f., Burr and Radin, 2003; Hayami et al., 2004). However, if density changes are significant for the etiology of osteoarthritis, one could ask whether the bone density must increase beyond a certain level before cartilage deterioration would result, and also, what alterations to the normal loading patterns are necessary to produce clinically significant changes in bone density?

Overall, this study demonstrates that analyses of the spatial patterns of subchondral apparent density can be used to reconstruct experimentally induced differences in joint posture, and these results should permit the reconstruction of habitual posture for osteological specimens of extant taxa, particularly for those in museum collections, as well as for taxa represented by subfossilized skeletal remains (Polk et al., 2006). However, we must add a few caveats to the results that we have obtained. First, the technique of analyzing patterns of subchondral apparent density may not be useful for reconstructing posture at all joints. Factors such as the range of segmental movement or joint shape (e.g., congruency or breadth, etc.) may limit variation in the distribution of loads on the opposing joint surfaces. Ahluwalia (2000), for example, has observed that primates did not exhibit systematic taxonomic variation in the anterior–posterior position of the RMD on the tibial plateau. This result may be influenced by the fact that the shank range of motion does not typically vary as much as the thigh (Kadaba et al., 1995), and because the articular surface on the tibial plateau is relatively limited in size. In this study, we analyzed only on a single slice through the MFC. This is appropriate for a joint where the condyles are relatively narrow and the joint acts primarily as a hinge. We have not examined systematically any variation in the position of the RMD across the joint surface but for broader joints this type of analysis would be warranted (c.f., Carlson and Patel, 2006; Patel and Carlson, 2007).

Second, these results were obtained for a sample of juvenile animals whose bone may have acclimated more quickly to altered loading regimens than would those of adult animals. Bone remodeling rates decrease in older individuals (Forwood and Burr, 1993; Lieberman et al., 2003; Ruff et al., 2006) and for this reason the patterns of subchondral apparent density may reflect longer-term loading histories in adults. The duration and magnitude of the loading history required to effect a change in apparent density in adult animals remains to be tested in more detail. However, we assume that different patterns of subchondral density in adult individuals will permit inferences to be made about differences in their joint postures. Finally, given the large degree of variation that exists in the position of the RMD in our control subjects, we recommend analyses of intraspecific variation before making interpretations about the significance of interspecific differences in these density measures.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED

All procedures related to animal care and use were approved by the University of Illinois IACUC #04289. Tom Nash and Dick Cobb helped to coordinate sheep acquisition, transport, housing, and feeding. Jessica Travers, Jeff Peterson, Kyle Johnson, Mark Boyer, and Gianni Pezzarrossi assisted with the training and exercise of the sheep. UIUC veterinary staff provided excellent care. Kevin Reynolds, at Mount Auburn Hospital in Cambridge, and the staff in cross-sectional imaging provided exceptional service and assistance with all of the CT scanning. Daniel Weber, Janet Hanlon, and Petra Jelinek assisted with training on AMIRA at the Visualization, Media and Imaging Laboratory at the Beckman Institute for Advanced Science and Technology. Nathan Young gave excellent advice for obtaining coordinate data. Undergraduate research assistance and data collection was provided by Jeff Peterson, Ashley Warmoth, Gianni Pezzarossi, Juliana Monsalve, Jim Vasa, and Katie Bott. This research was funded by the University of Illinois Research Board.

LITERATURE CITED

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED