Local Variation in Femoral Neck Cortical Bone: In Vitro Measured Bone Mineral Density, Geometry and Mechanical Properties

ARTICLE IN PRESS Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health, vol. ■, no. ■, 1–11, 2015 © Copyright 2015 by The International Society for Clinical Densitometry 1094-6950/■:1–11/$36.00 http://dx.doi.org/10.1016/j.jocd.2015.10.003

Original Article

Local Variation in Femoral Neck Cortical Bone: In Vitro Measured Bone Mineral Density, Geometry and Mechanical Properties Louise V. Coutts,*,1 Thomas Jenkins,1 Richard O. C. Oreffo,1 Doug G. Dunlop,2 Cyrus Cooper,3 Nicholas C. Harvey,3 Philipp J. Thurner,1,4 and the OStEO Study Teama 1

Faculty of Engineering and Environment and Faculty of Medicine, University of Southampton, Southampton, UK; Southampton University Hospitals NHS Trust, Southampton, UK; 3MRC Lifecourse Epidemiology Unit, University of Southampton, Southampton, UK; and 4Institute of Lightweight Design and Structural Biomechanics, TU Wien, Vienna, Austria 2

Abstract Age- and disease (osteoporotic fractured and osteoarthritic tissue)-related changes in the distribution of cortical bone were examined, using a multimodality approach, including measurement of local density, geometry and mechanical properties, where changes in these properties can give rise to instability and increasing probability of fracture. In contrast to the majority of previously reported research, this study also focuses on the characteristic non-circular femoral neck cross-sectional geometry and variation in bone mineral density (BMD) around the femoral neck. Twenty-two osteoarthritic and 7 osteoporotic femoral neck slices, collected from elective and trauma-related arthroplasty, and 16 cadaveric donor tissue controls were tested mechanically using Reference Point Indentation (BioDent™, Active Life Technologies®, Santa Barbara, CA) and then scanned with in vitro-based radiography intended to replicate the dual-energy X-ray absorptiometry technique. All parameters were measured regionally around the circumference of the femoral neck, allowing examination of spatial variability within the cortical bone. Fractured tissue was less resistant to indentation in the thinner superolateral segment compared to other segments and other groups. BMD around the fractured femoral necks appeared more consistent than that of nonfractured tissue, where BMD was reduced in the superolateral segment for the other groups. Cortical bone was thin in the superolateral segment for all groups except for the osteoarthritic group, and was thicker in the inferomedial segment for both osteoarthritic and fractured groups, resulting in the largest variation in buckling ratio (ratio of cortical bone diameter to cortical bone thickness) around the femoral neck for the fractured group. With age, healthy controls appeared to have lower inferomedial cortical thickness, whereas no significant differences in Reference Point Indentation measurements and density were observed. The study has highlighted several (both quality- and quantity-related) parameters that may be used to improve prediction of fracture risk. Key Words: Bone mineral density; cortical bone; indentation; mechanical properties; osteoporosis.

Introduction The ability of cortical bone tissue to withstand loading tends to deteriorate with age, where an imbalance between bone resorption and replenishment arises (1), increasing the susceptibility of bone to fracture. At present, this fracture risk is assessed using lifestyle measures with tools such as the Fracture Risk Assessment Tool and bone mineral density

Received 07/27/15; Revised 10/12/15; Accepted 10/21/15. a OStEO Study Team: Ball, Carole, Chan, Kitty, Taylor, Patricia, Katsamenis, Orestis, Latham, Jeremy and Arden, Nigel. *Address correspondence to: L. V. Coutts, PhD, Faculty of Engineering and Environment and Faculty of Medicine, University of Southampton, Southampton, UK. E-mail: [email protected]

1

ARTICLE IN PRESS 2 (BMD), measured with techniques such as dual-energy X-ray absorptiometry (DXA) (2–4).These measures highlight those with predetermined osteoporotic levels of particular concern to clinicians and often lead to prescription of medications such as bisphosphonates. However, although these diagnostic tools are good indicators for risk of fracture on the population scale, they are limited in accuracy on a case-by-case basis (5). With the recent availability of novel mechanical and imaging tools, localized assessment of structural and compositional bone properties that determine bone quality may be conducted alongside and complimentary to bone quantity measurements. Bone quantity is commonly measured in the clinic using techniques such as DXA, which measures BMD within intact bone in a 2-dimensional manner, effectively flattening the tissue, measuring a single mineral value per pixel, as the X-ray beam passes through the tissue. Hence, while the technique excludes the collagen component of bone as well as soft tissue, BMD measurements include both the cortical and trabecular bone that the X-ray beam has traveled through.Three-dimensional imaging techniques such as computed tomography have been used to analyze bone geometry in more detail around the circumference of the femoral neck, showing that within the femoral neck, especially in females, cortical bone geometry becomes increasingly radially inconsistent with age (6–8), where regions that typically undergo the lowest stresses typically decrease in thickness and BMD with age, increasing susceptibility to fracture if the patient were to fall and abnormally load the bone. To further separate low- and highrisk fracture groups, it would appear pertinent to assess more local changes in both bone geometry and structure than are currently undertaken on a routine clinical basis. Techniques that extract more localized information are therefore in demand for scientific understanding, possible future diagnosis, and assessment of treatment. Although ideally direct measurement of mechanical properties of cortical bone is preferable, instead of surrogate measures such as BMD, historically this has required removal of the bone, which is not practical for clinical use and undertaking methods such as strength or fracture toughness testing. Nanoindentation (9) and microindentation (10) methods have recently been developed to enable measurement of the mechanical properties of bone on a smaller scale. Surface methods able to be performed in vivo have recently been explored, such as the OsteoProbe and the method used in the present study: BioDent™ Reference Point Indentation (RPI) (Active Life Technologies®, Santa Barbara, CA), a new minimally invasive microindentation technique that involves making a small incision through the skin and then creating a small amount of damage in the superficial layer of cortical bone to measure the mechanical properties of cortical bone.These techniques allow direct measurement of mechanical properties and are currently undergoing proper evaluation prior to clinical use (11–16). In the case of the BioDent, a 375-μm-wide indenter makes successive indents while recording load and displacement

Coutts et al. to measure mechanical properties of bone (13–16). The RPI device consists of an outer reference probe that rests on the bone surface and an inner test probe that indents the bone while its movement is recorded relative to the reference probe. The reference probe allows the absolute indentation distance to be calculated. The test probe indents perpendicular to the bone surface, and it is proposed that the depth of these indents corresponds to fracture resistance; the more easily the bone is fractured, the further the test probe will indent the bone. While investigation is currently being undertaken to assess this hypothesis, such techniques nevertheless have potential to improve accuracy of fracture incidence prognosis by complementing existing methods such as use of single, whole-region BMD measurements alone. Although bone fracture risk is likely predominantly dependent on mechanical properties, it is also to an unknown extent dependent on factors such as bone quantity, as currently measured clinically by DXA. To allow direct comparison of bone quality and quantity, DXA was performed in vitro in this study, on slices of femoral neck bone, allowing localized cortical bone BMD and geometry around the femoral neck to be compared (on a site by site basis) with the direct measurement of mechanical properties with RPI. The study aimed to investigate trends in cortical bone geometry, BMD, and mechanical properties around the femoral neck, in fractured and nonfractured bone, to further investigate differences between tissue types and to improve understanding of susceptibility to fracture.

Methods Replication of the DXA Technique To overcome hygiene restrictions and to achieve improved image resolution, samples were not scanned using a clinical DXA device; instead the DXA technique was replicated by applying the same measurement principle to images acquired using a 225-kV X-Tek HMX ST tomography device (Nikon Metrology Europe NV, Heverlee, Belgium) in radiography mode. This method allowed trends around the femoral neck to be investigated at higher resolution than standard clinical DXA. The aim of the study was to inform future development of clinical DXA by providing high-resolution information on trends in femoral neck properties, rather than developing a replicable novel technique for conversion of radiographs to absorption images. Differences in data output between the conversion technique and current clinical DXA were minimized by undertaking the calibration procedure detailed at the end of this section. DXA is based on the Beer–Lambert law, where I is image intensity, Io is initial intensity prior to attenuation, μ/ρ (or μm) is the mass attenuation coefficient and ρl is the areal density (in grams per square centimeter):

Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health

I = Ioe

⎛μ ⎞ − ⎜ ρl ⎟ ⎝ρ ⎠

(1) Volume ■, 2015

ARTICLE IN PRESS Localized BMD, Geometry and Mechanical Properties for Improved Fracture Risk Prediction The equation is expanded to include each of the 2 tissue types: soft tissue (st) and bone tissue (b), and duplicated for each energy, allowing attenuation of the X-ray beam to be solved for each tissue type:

I = I oe

⎛⎛ μ ⎞ ⎛ μ ⎞ ⎞ −⎜ ⎜ ρl ⎟ + ⎜ ρl ⎟ ⎟ ⎝ ⎝ ρ ⎠ st ⎝ ρ ⎠ b ⎠

(2)

The μ/ρ (or μm) of bone is well documented for each input X-ray energy (17); however, the μm of soft tissue varies from person to person, depending on the ratio of lean to fat tissue, but can easily be calculated from regions where the X-ray has passed through only soft tissue. Equation (1) is used with soft tissue parameters only (as no bone is involved) for both low (L) and high (H) energies and combined (as the areal density [ρl] of soft tissue will not change between acquisitions at different energies) and then rearranged to give the ratio of μ m for soft tissue, R st as

μ Rst = mLst = μmH st

⎛I ⎞ ln ⎜ L ⎟ ⎝ I oL ⎠ ⎛I ⎞ ln ⎜ H ⎟ ⎝ I oH ⎠

(3)

With Rst for each sample known, areal density of bone may be calculated. Initial intensity and image intensity are available from pixel values within radiographs; therefore 2 unknowns remain in Eq. (2) (ρlst and ρlb). To solve for these 2 remaining unknown values, 2 versions of Eq. (2) are therefore required, hence the requirement for imaging at 2 different energies, fulfilling Eqs. (4) and (5):

I L = I oL ⋅ e

I H = I oH ⋅ e

⎛⎛⎛ μ⎞ ⎞ ⎞ ⎞ ⎛⎛ μ⎞ −⎜ ⎜ ⎜ ⎟ ρl ⎟ + ⎜ ⎜ ⎟ ρl ⎟ ⎟ ⎝ ⎝ ⎝ ρ ⎠ L ⎠ st ⎝ ⎝ ρ ⎠ L ⎠ b ⎠

(4)

⎛⎛⎛ μ⎞ ⎞ ⎞ ⎞ ⎛⎛ μ⎞ −⎜ ⎜ ⎜ ⎟ ρl ⎟ + ⎜ ⎜ ⎟ ρl ⎟ ⎟ ⎝ ⎝ ⎝ ρ ⎠ H ⎠ st ⎝ ⎝ ρ ⎠ H ⎠ b ⎠

(5)

As for when calculating Rst for soft tissue, the areal density of soft tissue (ρlst) will be constant for both energy images; therefore, Eqs. (4) and (5) can be combined and rearranged to give the equation for the areal density of bone (measured in grams per square centimeter):

⎛I ⎞ ⎛I ⎞ ln ⎜ H ⎟ ⋅ Rst − ln ⎜ L ⎟ ⎝ I oH ⎠ ⎝ I oL ⎠ ρlb = μmLb − ( μmLb ⋅ Rst )

(6)

The density measured in each pixel (areal density) will include the entire depth of the sample which the X-ray beam has passed through; that is, a thick sample will yield higher areal density values than a thin sample. Femoral neck slices were

3

scanned alongside an 8-mm-diameter nylon cylindrical rod (present for all DXA replica scans) to allow calibration of areal density measurements.The rod enabled measurement of beam attenuation through an object of known μm and highly varying thicknesses across the curved surface of the cylinder, providing a large range of different areal densities, from the thinnest part at the edge of the curved face to the thickest part at the center. Rod geometry and material were selected so that the resulting areal density range matched that expected in the sample being imaged. The entire image could then be recalibrated based on the relationship between the rod image intensity values and the predicted image intensity values [calculated for each areal density using Eq. (1)]. Pixel intensities imaged at matching thicknesses, along the length of the cylindrical rod, were first averaged to minimize image noise, resulting in the typical average measured intensity shown in Fig. 1a.As the material density and μm were known, the local areal density (density × thickness) could be calculated for each point across the curved radius of the cylindrical rod.This provided multiple points for a linear fit between the average measured pixel intensity values and the calculated pixel intensity values, using the least squares method. On average, for both high- and low-energy images, the least squares linear fit between the measured intensities (I) and the calculated intensities (Ical) was

I cal = 6 I − 5

(7)

Pixel intensities for the entire low- and high-energy images were then recalibrated producing calibrated intensity images. Calibrated intensity values in the region of the cylindrical rod are shown in Fig. 1a. Density was then calculated on a pixel-by-pixel basis as described in the previous section, using the calibrated image intensities, providing BMD images of each sample. To simulate a Hologic Discovery A clinical DXA scanner (Hologic, Inc., Bedford, MA), the above method was applied to images acquired of all femoral samples at dual energies, using the following settings: peak voltages 100 and 140 keV, exposure time 1415 ms and current 50 μA. To eliminate movement of samples between images, the same 1-mm Cu filter was used to reduce beam-hardening effects in the sample for all images; hence the Hologic spectrums were not matched explicitly. The setup resulted in an approximate resolution of 26 μm/pixel. An example of a femoral neck slice density image is shown in Fig. 2. To validate the calibration technique, 6 whole proximal femur samples (before neck slices were cut) were also scanned, using a 12.5-mm-diameter aluminum cylindrical rod for calibration. The total mass (total bone mineral, lean and fat content) correlated well with the physical mass of the sample (as measured with a scale) (r = 0.98, Fig. 1b). To ensure the 2 X-ray systems produced correctly calibrated density data, the aluminum calibration bar was also scanned with the Hologic Discovery A, where each repeat scan measured the total mass of the aluminum bar

Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health

Volume ■, 2015

ARTICLE IN PRESS 4

Coutts et al.

Fig. 1. BMD calibration: (a) image intensities as calculated, measured, and calibrated, against the varying thicknesses of the cylindrical calibration bar and (b) comparison of masses of 6 cadaveric proximal femoral neck sections: actual (measured with a scale) against densitometry calculated mass (bone and soft tissue inclusive). BMD, bone mineral density.

Fig. 2. In vitro osteoarthritic sample DXA replica density image (superolateral section missing), in grams per square centimeter (white = high density, black = low density). Pixel size 0.026 × 0.026 mm. DXA, dual-energy X-ray absorptiometry.

to within 0.2 g of the actual mass of the bar (as measured with a scale).

Tissue Samples Twenty-two osteoarthritic (age range: 26–84) and 7 osteoporotic (age range: 70–89) femoral neck samples from hip fracture surgery following elective arthroplasty or trauma at Southampton University Hospital NHS Trust were acquired from both male and female subjects, who gave

informed consent. Sixteen disease-free whole femur cadaveric donor samples (age range: 57–92, free from bone disease) were also sourced as a control from Innoved Institute LLC (Bensenville, IL—15 samples) and International Institute for the Advancement of Medicine (Edison, NJ—1 sample). Full ethical approval was granted under application numbers 12/SC/0325 and 10/H0604/91. Sevenmillimeter-thick femoral neck slices were then cut (mean slice thickness standard deviation within slices was 0.53 mm) by the investigators using a handsaw. Osteoarthritic and

Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health

Volume ■, 2015

ARTICLE IN PRESS Localized BMD, Geometry and Mechanical Properties for Improved Fracture Risk Prediction Proximal Cut Neck Slice Distal Cut

5

whole intact femurs; hence, for the control samples, both proximal and distal cuts were made by the investigators. The fragility of fractured samples resulted in (1) 2 samples missing the inferomedial section and (2) 1 further neck slice sample was untestable with RPI due to inability of the tissue to support the indentation load; hence, for this region, imagebased measurements were averaged over only 5 samples and RPI over only 4 samples.

Image Processing Femoral Neck

Fig. 3. Proximal femur diagram, indicating the orientation and location of femoral neck slice samples. osteoporotic samples were provided from surgery such that the distal cut of the femoral neck slice (see Fig. 3) was performed by the surgeon and the proximal cut was made by the investigators. Cadaveric samples were provided as

Cortical bone inner and outer boundaries were selected manually from density images, using the MATLAB (Natick, MA) “createMask” function. Circumferential variations in the following parameters around the femoral neck were then automatically calculated using a custom MATLAB algorithm, by averaging the following on a degree-by-degree basis: cortical bone density, cortical bone thickness, cortical diameter (calculated as the average of the inner and outer cortical diameters), and ratio of the outer cortical diameter (on a degree-by-degree basis) to the cortical bone thickness (buckling ratio; see Fig. 4). The centroid was calculated as the mean coordinate of the inner and outer cortical bone surfaces. Radial variation in density through the depth of the cortical bone was also analyzed INFERO_MEDIAL Cortical Thickness

INFERO_MEDIAL

Outer Cortical Edge

Inner Cortical Edge

POSTERIOR

ANTERIOR

POSTERIOR

ANTERIOR

Outer Cortical Diameter

SUPERO-LATERAL

Inner Cortical Diameter

SUPERO-LATERAL

Fig. 4. (a) In vitro osteoarthritic sample DXA replica density image (grams per square centimeter; white = high density, black = low density). Pixel size 0.026 × 0.026 mm. (b) Corresponding example schematic of measurements conducted on femoral neck slices. DXA, dual-energy X-ray absorptiometry. Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health

Volume ■, 2015

ARTICLE IN PRESS 6

Coutts et al.

by rescaling by cortical thickness on a degree-by-degree basis, and averaging circumferentially through anterior, inferomedial, and posterior regions, with the thin, more fragile (and consequently sometimes missing; see Fig. 2) superolateral region excluded from the analysis.

RPI To investigate the heterogeneity in mechanical properties, 18 mechanical indentation tests were performed around the circumference of the femoral neck slices using the BioDentTM RPI instrument, into the surface of the bone in the radial direction. For each indentation test, 10 indentation cycles were applied at 2 Hz, each with a maximum force of 10 N. Each loading cycle consists of 3 equal time periods. During the first period, the load is applied linearly; during the second, the load is held; and during the third, the load is removed, also in a linear manner. For the entire period, the distance of the test probe from the reference probe is continually recorded (indentation distance), allowing various parameters to be assessed such as initial indentation distance, indentation distance increase during the hold time (creep), and between consecutive cycles.

Statistical Analysis Where parameters were normally distributed across donors, arithmetic means were calculated and 2-tailed t tests and Pearson’s correlations were performed. Mann-Whitney U-tests were undertaken for non-normally distributed parameters. Statistical analysis was completed using SPSS v20 (IBM, Portsmouth, UK).

Results Table 1 shows the mean (or median for non-normally distributed variables) and standard deviation per group summaries for the measured parameters. For each of the 18 circumferential test sites, the indentation distance increase

across all cycles (IDI) and the first cycle creep (CID) across all femoral neck slices within each group are plotted in Fig. 5. Average cortical bone density (grams per square centimeter) is shown (1) around the circumference of the femoral neck slices, calculated on a degree-by-degree basis and averaged across each of the groups (Fig. 6a); and (2) through the depth of the cortical femoral neck bone (Fig. 7). For each neck slice, BMD was normalized to the BMD in the inferomedial region, to reduce variability arising from different neck slice thicknesses. The superolateral sections have lower cortical BMD than the inferomedial sections, except for the fractured group (see Fig. 6a). On a case-by-case basis, an approximately linear trend can be observed between the 2 regions. Therefore, the rate of change of density around the femoral neck was quantified for each sample as the linear least squares fit gradient of the reduction in density (by degrees) around 120° of the neck circumference from both (1) the inferomedial region to the anterior side and (2) the inferomedial region to the posterior side. As anterior and posterior side gradients were similar, a mean value across both sides was then computed as shown in Fig. 6b. The median gradients (and corresponding correlations) per group are shown in Table 1. As the BMD in the fractured group did not follow a similar pattern around the femoral neck to that of the osteoarthritic and control groups, when linear trends were compared on a case-by-case basis, fractured and control groups were significantly different (Mann-Whitney U-test for nonnormal distribution, p < 0.05). The average cortical bone thickness (millimeter) and the ratio of the outer cortical bone diameter to the cortical bone thickness (buckling ratio) around the femoral neck are plotted in Fig. 8a and c, respectively, and averaged across each of the groups. Figure 9 shows the cortical bone thickness, IDI, and density in the inferomedial region against body mass and age. For the control group, some positive correlation appeared between thickness and body mass (r = 0.60, p = 0.014), and for the fractured group, some positive correlation appeared between IDI and body mass

Table 1 Group Average (and Standard Deviation) by Measured Parameter, Averaged Over the Entire Circumference Unless Otherwise Specified as IM, SL, A, or P Parameter Cortical thickness (IM) Buckling ratio (IM) Buckling ratio (SL) RPI: IDI RPI: CID Ratio of IM–SL : A–P femoral neck diameter Circumferential rate of change in density g/ (°.cm2) and corresponding correlation, r

Osteoarthritic 4.08 mm 4.08 7.88 26.4 μm 11.6 μm 1.78 0.024

(1.70 mm) (0.21) (0.70) (2.5 μm) (1.2 μm) (0.33) (r = 0.82)a

Control 3.19 mm 5.94 15.74 16.8 μm 7.5 μm 1.56 0.020

(1.20 mm) (0.24) (1.29) (2.4 μm) (0.9 μm) (0.14) (r = 0.84)a

Fractured 6.13 mm 2.97 14.31 25.5 μm 11.3 μm 2.19 −0.002

(1.53 mm) (0.06) (1.84) (8.3 μm) (3.0 μm) (0.57) (r = 0.48)a

Abbr: A, anterior; CID, first cycle creep indentation distance; IDI, indentation distance increase; IM, inferomedial; P, posterior; RPI, Reference Point Indentation; SL, superolateral. a Non-normally distributed therefore median calculated. Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health

Volume ■, 2015

ARTICLE IN PRESS Localized BMD, Geometry and Mechanical Properties for Improved Fracture Risk Prediction

7

Fig. 5. (a) IDI and (b) CID in microns, in the radial direction around femoral neck slice samples, averaged across subjects for each group. CID, first cycle creep indentation distance; IDI, indentation distance increase.

Fig. 6. (a) Mean cortical bone density (grams, per square centimeter) normalized to BMD in the inferomedial region of the femoral neck slice sample, averaged across groups (bold) and for linear fit regions (between the vertical lines), 1 sample case for each group (faint). (b) Linear reduction in BMD around the femoral neck (from the inferomedial to the superolateral regions). BMD, bone mineral density. Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health

Volume ■, 2015

ARTICLE IN PRESS 8

Coutts et al.

Fig. 7. Mean BMD through the radial (depth) direction, averaged around the femoral neck (excluding superolateral region) and normalized to average overall neck slice BMD. BMD, bone mineral density.

(r = 0.81, p = 0.19), between thickness and age (r = 0.55, p = 0.34) and between density and IDI (r = 0.88, p = 0.12), although none of these correlations reached significance. Figure 8b shows the diameter of the femoral neck from the mid-depth of the cortical bone to the mid-depth of the cortical bone on the opposite side. There is a distinctive narrowing across the femoral neck in the anterior–posterior direction for all groups, even more so for both fractured and osteoarthritic groups. Figures 10 and 11 show the profiles of BMD through the femoral neck for cortical bone, both including and excluding trabecular bone, respectively, from inferomedial to superolateral regions, reconstructed in the same orientation as acquired for a clinical BMD scan.

Discussion Measurable differences between control and fractured tissue may have potential to improve fracture risk assessment in those with suspected osteoporosis. In addition, with the known reduction in risk of femoral neck fracture in osteoarthritis, any differences between the fractured and the osteoarthritis group may provide further understanding of mechanisms leading to fracture. All parameters displayed similar trends around the circumference of the femoral neck, albeit to different magnitudes, except for BMD and RPI measurements in fractured tissue. Differences within both the inferomedial and superolateral regions were expected as these are the regions undergoing high stress through normal gait and in the reverse compressive/tensile manner through high-impact fall loading (18), where abnormally large compression is applied to the superolateral region. The observed trends are in agreement with typical loading patterns around the femoral neck where highest stresses, arising from normal gait, can typically be found in the

Fig. 8. (a) Mean cortical bone thickness (millimeter) around femoral neck slice samples; (b) femoral neck bone diameter (measured from the mid-depth of the cortical bone) around the neck slice samples; (c) mean ratio of the outer cortical bone diameter to the cortical bone thickness (buckling ratio) around the circumference of femoral neck slice samples, all averaged across each group. inferomedial region. The study has shown that this region has the highest cortical thickness, the highest femoral neck diameter, and the lowest buckling ratio for all groups, the highest cortical BMD for the control and osteoarthritic

Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health

Volume ■, 2015

ARTICLE IN PRESS Localized BMD, Geometry and Mechanical Properties for Improved Fracture Risk Prediction

9

Fig. 9. Inferomedial region: (a) cortical bone thickness (millimeter); (b) IDI; (c) density, displayed per subject, against body mass and age; (d) density against cortical bone thickness; (e) IDI against cortical bone thickness; and (f) density against IDI. Displayed density values were normalized to the entire neck slice. Linear correlations greater than r = 0.5 are shown. IDI, indentation distance increase. groups, and for the fractured group, also the lowest RPI measurements. Cortical bone thickness follows a similar trend around the femoral neck for all diseases. In the thickest, inferomedial region, thickness tends to be largest for the fractured group, compared to both the osteoarthritic group and the control group (see Table 1), where the thickness in the most inferomedial point is significantly different between fractured and control group samples (p < 0.03). This effect is also apparent in the DXA data (see Fig. 11) where the width of the initial peak reflects the thickness of the cortical bone in the inferomedial region. Therefore, while for the fractured group the cortical bone in the inferomedial region is thicker, the commonly found “empty” or porous regions within it lead to overall reduced BMD

in this group. However, as these porous regions are effectively averaged out for typical clinical DXA scanning (or flattened due to imaging only in 2 dimensions), using density data for cortical thickness measurement may have limited accuracy. The correct setting of boundary thresholds is important in order not to confuse cortical bone porosity with the transition zone between cortical and trabecular bones. In the superolateral region, cortical thickness is lowest for the control and fractured groups. Greater variation in thickness for both osteoarthritic and fractured groups in this region is perhaps indicative of small pockets of thin cortical bone and deformities. Such thin regions may weaken the structure, allowing cracks to propagate more easily than in healthy cortical bone, where changes in cortical thickness are more gradual. Thicker cortical bone in this region

Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health

Volume ■, 2015

ARTICLE IN PRESS 10

Fig. 10. Mean cortical and trabecular bone BMD through the femoral neck, normalized to average slice cortical BMD, reconstructed in the same orientation as acquired for a clinical BMD scan. BMD, bone mineral density.

for the osteoarthritic group may be responsible for the reduced susceptibility to femoral neck fracture in osteoarthritis sufferers and may arise from altered gait to reduce pain. Similar results have previously been reported, where cortical bone in the superolateral region was found to be thicker in osteoarthritic bone (19–22) and thinner in osteoporotic bone (8,19,22,23), leading to increased fragility in a fall scenario, where the superolateral region is compressed, and the thin cortical bone may buckle and initiate fracture. Fracture initiation is more likely where the femoral neck diameter is more elongated, as previously observed in osteoporosis, whereas the opposite has been found to occur in osteoarthritis, where the femoral neck is rounder (19).

Fig. 11. Mean BMD of cortical bone (trabecular bone excluded) through the femoral neck, normalized to average slice BMD, reconstructed in the same orientation as acquired for a clinical BMD scan. BMD, bone mineral density.

Coutts et al. For this study, the mean inferomedial–superolateral to anterior–posterior ratio of the femoral neck diameter was significantly different between the fracture group and the control group (p < 0.01). However, when the groups were compared, the osteoarthritic group ratio was not lower than the control group (i.e., not rounder in shape) as previously reported; some individual osteoarthritic samples had a lower ratio than all control samples. Femoral neck diameter is typically largest in the inferomedial–superolateral direction.A typically thinner cortical bone in the superolateral region results in a higher outer cortical bone diameter-to-thickness (buckling) ratio in the superolateral region. The thicker cortical bone measured in the osteoarthritic group in the superolateral region results in a significantly lower buckling ratio compared to the other groups (see Table 1; p < 0.05).The control and fractured group trends appear similar except for in the inferomedial region, where buckling ratio is extremely low for the fractured group. Comparing buckling ratio in the inferomedial region on an individual basis yielded significant differences between groups (see Table 1; p < 0.05). Hence, comparing differences in buckling ratio between inferomedial and superolateral regions may be a useful quantifiable feature determining susceptibility to fracture, a feature that is possible to calculate from current clinical DXA scans, in a similar manner to that currently used for the single-measurement buckling ratio in Hip Structure Analysis. For RPI, both osteoarthritic and fractured groups had larger IDI and CID compared with the control group (see Table 1; p < 0.05). While osteoarthritic and fractured group data were similar in magnitude, when circumferential test data were split into 2 groups by test position, for the fractured group, the mean IDI for the superolateral half was 5.38 μm greater than the inferomedial half, yet for the osteoarthritic group, the mean IDI in the superolateral half was 0.78 μm less than the inferomedial half. These group differences are merely indicative of possible differences with condition as they were found to be nonsignificant, possibly due to the small number of samples or high variation in RPI measurements. The authors are currently undertaking a further study with a larger sample size to further validate these findings. The lower cortical BMD observed in the superolateral regions compared to the inferomedial regions, in the control and osteoarthritic groups, but not in the fractured group, highlights a further potential measurement suitable for improved fracture risk prediction. In addition, although not separately reported, trends were also similar when considering just the outermost section of cortical bone, where RPI measurements were conducted. In conclusion, the study has highlighted several parameters that may be used for fracture risk prediction, including those based on geometry, mechanical properties, and BMD, with some of these measurement data already being available through currently used clinical techniques. The inferomedial region of the femoral neck appears to show the most interesting differences between the fractured and

Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health

Volume ■, 2015

ARTICLE IN PRESS Localized BMD, Geometry and Mechanical Properties for Improved Fracture Risk Prediction control groups; however, care should be taken that these differences are not related to other diseases such as osteoarthritis. More localized assessment of femoral neck bone appears to be possible and worthwhile, for both fracture risk assessment and for assessing response to treatment of osteoporosis.

Acknowledgments We gratefully acknowledge funding of this research by an Engineering and Physical Sciences Research Council (EPSRC) grant (EP/J008192/1).Additional support was provided by a studentship partly funded by University of Southampton alumnus Mike Russell, the Medical Research Council, NIHR Musculoskeletal Biomedical Research Unit, University of Oxford, and NIHR Biomedical Research Centre, University of Southampton and University Hospital Southampton NHS Foundation Trust. We acknowledge μ-VIS X-ray imaging center at the University of Southampton for provision of tomographic imaging facilities. We additionally show our appreciation to the Osteoporosis Centre, University Hospital Southampton NHS Trust for performing the bone mineral density scans of the osteoporotic groups. Finally, we thank all those involved in OStEO (Observational Study Examining Osteoporosis) and, importantly, the participants of this research study.

References 1. Currey JD. 1979 Changes in the impact energy absorption of bone with age. J Biomech 12(6):459–465. 2. Siris ES, Chen YT, Abbott TA, et al. 2004 Bone mineral density thresholds for pharmacological intervention to prevent fractures. Arch Intern Med 164(10):1108–1112. 3. Schuit SCE, van der Klift M, Weel AEAM, et al. 2004 Fracture incidence and association with bone mineral density in elderly men and women: the Rotterdam Study. Bone 34(1):195–202. 4. Humadi A, Alhadithi RH, Alkudiari SI. 2010 Validity of the DEXA diagnosis of involutional osteoporosis in patients with femoral neck fractures. Indian J Orthop 44:73–78. 5. Marshall D, Johnell O, Wedel H. 1996 Meta-analysis of how well measures of bone mineral density predict occurrence of osteoporotic fractures. BMJ 312:1254–1259. 6. Johannesdottir F, Poole KES, Reeve J, et al. 2011 Distribution of cortical bone in the femoral neck and hip fracture: a prospective case-control analysis of 143 incident hip fractures; the AGES-Reykjavik Study. Bone 48(6):1268–1276. 7. Poole KES, Mayhew PM, Rose CM, et al. 2010 Changing structure of the femoral neck across the adult female lifespan. J Bone Miner Res 25(3):482–491.

11

8. Mayhew PM, Thomas CD, Clement JG, et al. 2005 Relation between age, femoral neck cortical stability, and hip fracture risk. Lancet 366:129–135. 9. Rho J-Y, Tsui TY, Pharr GM. 1997 Elastic properties of human cortical and trabecular lamellar bone measured by nanoindentation. Biomaterials 18(20):1325–1330. 10. Coats AM, Zioupos P, Aspden RM. 2003 Material properties of subchondral bone from patients with osteoporosis or osteoarthritis by microindentation testing and electron probe microanalysis. Calcif Tissue Int 73(1):66–71. 11. Jenkins T, Coutts LV, Dunlop DG, et al. 2015 Site dependent reference point microindentation complements clinical measures for improved fracture risk assessment at the human femoral neck. J Bone Miner Res doi:10.1002/jbmr.2605. 12. Jenkins T, Coutts LV, Dunlop DG, et al. 2015 Variability in reference point microindentation and recommendations for testing cortical bone: maximum load, sample orientation, mode of use, sample preparation and measurement spacing. J Mech Behav Biomed Mater 42:311–324. doi:10.1016/j.jmbbm .2014.09.030; [Epub 2014 Oct 24]. 13. Hansma PK, Turner PJ, Fantner GE. 2006 Bone diagnostic instrument. Rev Sci Instrum 77. 14. Hansma P, Turner P, Drake B, et al. 2008 The bone diagnostic instrument II: indentation distance increase. Rev Sci Instrum 79. 15. Hansma P, Yu HM, Schultz D, et al. 2009 The tissue diagnostic instrument. Rev Sci Instrum 80. 16. Randall C, Mathews P, Yurtsev E, et al. 2009 The bone diagnostic instrument III: testing mouse femora. Rev Sci Instrum 80. 17. Pietrobelli A, Formica C, Wang ZM, Heymsfield SB. 1996 Dual-energy X-ray absorptiometry body composition model: review of physical concepts. Am J Physiol 271:941– 951. 18. Nawathe S, Nguyen BP, Barzanian N, et al. 2015 Cortical and trabecular load sharing in the human femoral neck. J Biomech 48(5):816–822. 19. Rubinacci A, Tresoldi D, Scalco E, et al. 2012 Comparative high-resolution pQCT analysis of femoral neck indicates different bone mass distribution in osteoporosis and osteoarthritis. Osteoporos Int 23(7):1967–1975. 20. Blain H, Chavassieux P, Portero-Muzy N, et al. 2008 Cortical and trabecular bone distribution in the femoral neck in osteoporosis and osteoarthritis. Bone 43(5):862– 868. 21. Neilson M, White A, Malik U, et al. 2004 Changes in bone architecture in the femoral head and neck in osteoarthritis. Clin Anat 17(5):378–391. 22. Boutroy S, Vilayphiou N, Roux JP, et al. 2011 Comparison of 2D and 3D bone microarchitecture evaluation at the femoral neck, among postmenopausal women with hip fracture or hip osteoarthritis. Bone 49(5):1055–1061. 23. Coutts LV, Jenkins T, Dunlop DG, et al. 2015 Variability in reference point microindentation and recommendations for testing cortical bone: location, thickness and orientation heterogeneity. J Mech Behav Biomed Mater 46:292– 304.

Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health

Volume ■, 2015