Differences in Sensitivity to Microstructure between Cyclic- and Impact-Based Microindentation of Human Cortical Bone

Abstract

Unlike the known relationships between traditional mechanical properties and microstructural features of bone, the factors that influence the mechanical resistance of bone to cyclic reference point microindention (cRPI) and impact microindention (IMI) have yet to be identified. To determine whether cRPI and IMI properties depend on microstructure, we indented the tibia mid-shaft, the distal radius, and the proximal humerus from 10 elderly donors using the BioDent and OsteoProbe (neighboring sites). As the only output measure of IMI, bone material strength index (BMSi) was significantly different across all 3 anatomical sites being highest for the tibia mid-shaft and lowest for the proximal humerus. Total indentation distance (inverse of BMSi) was higher for the proximal humerus than for the tibia mid-shaft but was not different between other anatomical comparisons. As a possible explanation for the differences in BMSi, pore water, as determined by ¹H nuclear magnetic resonance, was lowest for the tibia and highest for the humerus. Moreover, the local intra-cortical porosity, as determined by micro-computed tomography, was negatively correlated with BMSi for both arm bones. BMSi was however positively correlated with peak bending stress of cortical bone extracted from the tibia mid-shaft. Microstructural correlations with cRPI properties were not significant for any of the bones. The one exception was that average energy dissipated during cRPI was negatively correlated with local tissue mineral density in the tibia mid-shaft. With higher indentation force and larger tip diameter than cRPI, only IMI appears to be sensitive to the underlying porosity of cortical bone.

1. Introduction

Developed to improve the prediction of fracture risk^1,2,3, cyclic reference point microindentation (cRPI) and impact microindentation (IMI) are relatively new approaches to the characterization of mechanical properties of bone. Similar to traditional microindentation and nanoindentation techniques, cRPI and IMI involve measuring the depth of the indenter tip into a material, but whereas the traditional techniques require a sample with a flat surface, cRPI and IMI can be performed directly on the bone through the skin and periosteum of a patient⁴. To date, how cRPI and IMI properties relate to overall fracture resistance of bone is not well established and is likely dependent on the conditions affecting bone (e.g., aging, diabetes, osteogenesis imperfecta, treatment). In contrast, traditional microindentation is understood to assess the ability of a material to resist permanent or plastic deformation, and for metal alloys, hardness directly correlates with tensile strength⁵.

The indentation mechanism for cRPI and IMI are different. Manufactured by the same company (ActiveLife Scientific, Inc., Santa Barbara, CA), the BioDent instrument cyclically loads the bone between 0 N and a user-prescribed force (2 N to 10 N) at 2 Hz for a given number of cycles (typically 10 or 20), while the OsteoProbe generates a one-time impact force around 40 N. Both use a stainless steel, 90° conical-spherical tip for indentation with the radius of the OsteoProbe tip being larger (10 µm) than the radius of the BioDent tip (2.5 µm). The time to reach the maximum depth is 167 ms and 0.25 ms for the BioDent and OsteoProbe, respectively, such that loading rate of IMI is over 2,600 times greater than the loading rate of cRPI. As such, the IMI measurement did not correlate (p>0.05) with the various properties acquired from cRPI when paired indentation tests were done on cadaveric tibia mid-shafts at a clinically accessible site⁶. Thus, these microindentation techniques do not necessarily characterize the same material characteristics of bone. Regardless, IMI (OsteoProbe) is likely going to be the instrument of choice for clinical use as it is handheld and less cumbersome to use than the BioDent in which a stand is recommended for holding the device over the bone to reduce variance among neighboring indents⁷.

In the nascent cRPI literature, there are reports of indentation properties strongly correlating (rat)⁸, weakly correlating (human)^9,10, or not correlating (mouse)¹¹ with apparent-level material properties of cortical bone. As for IMI, there is no study to date that has directly tested whether the IMI measurement correlates with material properties of cortical bone, though one recent study indented cadaveric tibia mid-shafts and then tested machined cortical bone samples from the same quadrant in bending in order to compare the acquired properties to microstructure (porosity) and composition (mineral density)¹². Still to be elucidated are the mechanical characteristics of bone tissue that these indentation procedures are measuring¹³.

Studies of microindentation using human bones have usually focused on the antero-medial surface of the tibia mid-shaft due to its minimally invasive accessibility. However, intraoperative microindentation testing of other bones with the OsteoProbe to aid surgical decisions is conceivable, especially since ~14% of fracture fixations of the proximal humerus fail because the bone is of such poor quality that it cannot hold the screws over time¹⁴ and since osteoporosis is a risk factor for this loss of reduction^15,16. To further explore the relationships between indentation resistance properties and bone characteristics, we performed paired indentation tests on the cadaveric tibia mid-shaft (TMS), the proximal humerus (PH), and the distal radius (DR), all acquired from 10 elderly donors. As the loading mechanisms and depth of penetration differ between OsteoProbe (~200 µm) compared to BioDent (~70 µm), we hypothesized that indentation resistance of bone by IMI is more sensitive to the underlying intracortical microstructure than the cRPI.

2. Materials and Methods

2.1. Study Design

We acquired fresh cadaveric bone at 3 anatomical sites (tibia mid-shaft, distal radius, and proximal humerus) from 10 elderly donors (5 male: 85.4±6.0 yrs; 5 female: 88.6±6.2 yrs), through the Vanderbilt University Medical Center donor program (Nashville, TN). The dissected tissue from both the left and right side were wrapped in gauze, soaked in phosphate buffered saline (PBS), sealed in a plastic bag, and stored at −20 °C until analyzed.

2.2. Reference point microindentation techniques

With each hydrated bone firmly secured in a vice to avoid movement during the impact of the tip into the bone surface (Fig. 1a), IMI was performed at 8 locations (Fig. 1b) for both the left and ride side to determine that potential anatomical differences do not depend on side. In current clinical use of IMI, the non-dominant tibia is typically indented, but this would not be an option for intraoperative use of IMI. Each indentation site was spaced approximately 2 mm from the previous indent. For the DR and PH, the sequential indents started at the distal or proximal end, respectively, and these sites were chosen for their proximity to typical osteoporotic fractures. Performed by the same individual, the OsteoProbe (ActiveLife Scientific, Inc., Santa Barbara, CA) was held perpendicular to the bone surface (Fig. 1a), and then it was slowly pushed down until ~10 N, the force required to trigger the impact load. After indenting a bone (left or right), 5 additional indents were performed on a BMSi-100 reference material (polymethylmethacrylate or PMMA) that was secured within a rigid stainless steel block as provided by manufacturer. The indentation depth from impact at each site was indexed to the harmonic mean (HM) of the indentation depth into PMMA (Fig. 1a), thereby giving bone material strength index (BMSi = HM depth into PMMA per depth into bone). The order in which the 3 bones per donor were indented was random, and a unique tip was used for each donor. Following the indentation tests, the specimens were again wrapped in PBS-soaked gauze and stored at −20 °C until the next analysis.

Figure 1.

Locations of the impact (IMI) and cyclic (cRPI) indents. The cadaveric bones were secured in a vice for IMI, and the depth into bone was indexed to the depth into a PMMA standard (a). Each bone was indented 8 times as shown in schematic showing relative location of indents and µCT region of interest or ROI (c). Secured in a clamp for cRPI, each bone was indented again 8 times in neighboring region to generate a force vs. displacement curve from which properties were determined (b).

With each bone secured in a clamp on the base of the BioDent (ActiveLife Scientific, Inc., Santa Barbara, CA), cRPI was performed at sites (2 mm spacing) near the impact indents (Fig. 1b) while maintaining hydration with PBS (left only since BioDent is currently not intended for intraoperative use). Due to the irregular bone surfaces, cRPI sites do not exactly span the same region as the IMI sites of arm bones, but there was overlap in the axial direction (Fig. 1b). Using a BP2 probe assembly, each measurement, performed without preconditioning, involved 20 indentation cycles between 0 N and 10 N at 2 Hz and a dwell period between loading and unloading (Fig. 1c). Three probes were used, each indenting all anatomical sites of 3–4 donors. Computed with a custom MATLAB^® script (The MathWorks, Natick, MA), cRPI properties included: the maximum or total indentation depth (TID), the indentation distance between the first and last cycle (IDI), the average creep or distance traveled by the probe when subjected to the maximum load (AvgCID), the average energy dissipated per cycle (AvgED), the average loading slope (AvgLS) and unloading slope (AvgUS). The average properties were computed from the 3^rd to the 20^th cycle⁹.

2.3. Micro-computed tomography analysis (µCT)

In order to scan the region of the overlapping indentation sites using a high-resolution µCT scanner, a cortical bone strip (approximately 37.5 mm long and 5 mm wide with the thickness depending on the donor’s cortex) was extracted using a water-irrigated, diamond-embedded band saw (Model C-40, Gryphon Inc. USA, Sylmar, CA). The long axis of each strip was aligned with that of the scanner tube, and the region of interest (ROI) included the thicker portion of the cortex in each bone (Fig. 2a). The specimens, while immersed in PBS, were scanned using a polychromatic X-ray beam source having a peak voltage of 55 kVp and tube current of 200 µA (µCT50 scanner, Scanco Medical, Brüttisellen, Switzerland). With a 0.5 mm aluminum filter in place, beam hardening was further minimized using the manufacturer’s correction. The raw image slices were acquired with 500 projections per full rotation and 550 ms integration time in a 10.2 mm field-of-view. A total of 1632 slices at an isotropic voxel size of 10 µm provided a scan length of 16.3 mm for each bone. All scans were calibrated against a phantom of hydroxyapatite (HA) with densities varying between 0 mgHA/cm³ and 800 mgHA/cm³ to convert the attenuation coefficients into volumetric mineral density.

Figure 2.

Micro-computed tomography analysis of the cortical bone containing the indents. The long axis of each bone strip was scanned (a), and after imaging, specimens were extracted for three-point bending tests (TMS only) as well as bound and pore water measurements (blue box). To determine the local porosity and local mineral density, the evaluations were performed within rectangular regions (green contour) positioned below the indent throughout the image stack (b). The cortical thickness was manually measured (red arrows) for the proximal, mid-point, and distal slices.

After image reconstruction and ensuring the long-axis of the bone strip was vertically aligned, a rectangular contour (3.5 mm × 0.75 mm) was positioned near the periosteal surface (~350–500 µm) to assess local microstructure and density (Fig. 2b). For 2 thin PH samples, the rectangular contour was 3.5 mm × 0.38 mm in order to fit the box within the cortex. Using a Gaussian noise filter (sigma=1.0, support=2.0) and a global threshold (≥650.3 mgHA/cm³), the manufacturer’s software was used to calculate tissue mineral density (TMD) and apparent volumetric bone mineral density (avBMD) of the local region below the indents. Using a negative thresholding limit (−500 mgHA/cm³ – 650.3 mgHA/cm³), another script provided the local porosity (Ct.Po). Cortical thickness (Ct.Th) was manually determined using the on-screen linear measurement tool and calculated as an average of 9 measurements of the distance from the periosteal to endosteal surfaces (Fig. 2b) acquired from proximal, middle, and distal slices.

2.4. Biomechanical analysis

The cortical bone strips from the tibia mid-shafts were machined into beam specimens for flexural biomechanical testing. The machining process involved removing any excess trabeculae from the endosteal surface as well as the indents and then grinding the periosteal and endosteal surfaces to obtain a uniform shape. Briefly, after mounting the periosteal surface to a microscope slide (25 mm × 75 mm 2 mm plastic slide, EXAKT Technologies, Inc., Oklahoma city, OK) by applying a drop of cyanoacrylate adhesive (LOCTITE 406 PRISM, Loctite, Inc., Westlake OH) to each corner of the strip, the endosteal surface was ground (400 CS, EXAKT Technologies, Inc., Oklahoma city, OK) on a wet 1200-grit silicon carbide paper (WS flex 18C, Hermes Abrasives Ltd. Virginia Beach, VA). After sonicating in distilled water, the bone sample was carefully pried from slide using a razor blade to cut through the adhesive at each corner. Following the same procedure, the endosteal surface was then mounted on a new slide in order to produce parallel surfaces by additional grinding. The beam specimens had the following nominal dimensions: length = 26.5–37.0 mm, width = 4.0–4.5 mm, and thickness = 1.25–1.50 mm. With the endosteal side facing down on the two supports of a three-point (3pt) bending fixture (Model 2810-413, Instron, Norwood, MA), each hydrated specimen was monotonically loaded at 3 mm/min to failure using a universal material testing frame (DynaMight 8841, Instron, Norwood, MA). The span between the two supports was adjusted (between 25.0 mm and 30.0 mm) such that span-to-thickness ratio for the specimens was approximately 20. A 5.18 mm segment of each beam specimen, near the site of fracture, was imaged by µCT at an isotropic voxel size of 14.8 µm to determine the distance between surface and bending axis (C_min), cross-sectional area (Area), and the moment of inertia with respect to the axis of bending (I_min) (Supplemental Fig. 1). Because 4 specimens had rough surfaces (i.e., trabecularization of the endosteal surface was not fully removed; Supplemental Fig. 1), we generated another 4 beams from the contralateral strips acquired from the indentation region, as previously described, and repeated 3pt bending tests. Using custom MATLAB scripts (The Mathworks Inc., Natick, MA), the yield or peak moment (M), stiffness (δ), and work-to-fracture (W_f), adjusted for span, were determined from the force vs. displacement data to estimate modulus [δ × span³/I_min/48], strength [M × C_min/I_min], and span-adjusted toughness [3 × W_f/L/Area]. The yield point was determined by the 0.2% offset method (i.e., the intersection between the force vs. displacement curve and a line parallel to the initial slope but offset along the abscissa to 0.002 × L²/12/C_min.)

2.5. ¹H nuclear magnetic resonance (NMR) relaxometry

A 6 mm bone segment encompassing at least two indentation sites (Fig. 2a) was cut and assessed with proton NMR relaxometry. Exploiting the much faster transverse relaxation time of the protons of water molecules bound to the mineralized fibrils (T2 ~ 400 µs) compared to the protons of free water (T2 ~ 1ms–1s), ¹H-NMR quantifies the bulk amount of collagen bound water and pore water present in the specimen¹⁷. The bone specimen and a reference microsphere of water were inserted into an in-house, low proton, loop-gap-style, radiofrequency (RF) coil and placed in a 4.7T horizontal-bore magnet (Varian Medical Systems, Santa Clara, CA). Upon 90°/180° RF pulses of 10µs/20µs duration, Carr-Purcell-Meiboom-Gill measurements with 10,000 echoes were collected at 100 µs spacing, yielding data that were fitted with multiple exponential decay functions to generate a T₂ spectrum¹⁸. Integrated areas of collagen-bound water (T₂ = 160 µs to 1ms) and pore water (T₂ = 1 ms to 800 ms) were compared to the area of the reference sample (T₂ = 800 ms to 10 s) and converted into water volumes. Bound water (BW) and pore water (PW) volume fractions were obtained by normalizing the water content by the specimen volume as determined by Archimedes principle¹⁰.

2.6. Statistical analysis

To determine whether there were differences in bone properties among the anatomical sites, we used repeated-measures ANOVA with a Geisser-Greenhouse correction for unequal variance (Prism 6, GraphPad Software, Inc., La Jolla, CA). In the event that the data for any anatomical site did not pass the Shapiro-Wilk normality test, the non-parametric Friedman test was performed. For post-hoc comparisons (TMS vs. DR, TMS vs. PH, and DR vs. PH), either Holm-Sidak’s (parametric) or Dunn’s (non-parametric) test determined whether differences were significant at a family-wise significance level of 0.05. To test for significant relationships between the microindentation properties and microstructural features or material properties at the apparent-level, linear regression was performed on bootstrapped data (500 replicates to not violate the normality assumption). This bootstrapping does not affect the coefficient of determination (R²), correlation coefficient (r), slope, or y-intercept, but does give more conservative p-values than parametric linear regression analysis.

3. Results

3.1. Relative changes in indentation properties

Defined as the property value divided by the mean of the 8 indents for a given bone, relative BMSi did not systematically increase or decrease along the long axis of the bones (Fig. 3). However, the relative BMSi for the most distal and the most proximal indent of the radius and humerus, respectively, was significantly lower than the relative BMSi values at the other sites (Fig. 3). Therefore, for all subsequent comparisons involving the arm bones, these sites were excluded from the mean calculations (average of 7, instead of 8 indents). All relative cRPI properties from the BioDent did not significantly vary among the neighboring indent locations, regardless of anatomical site (Supplemental Fig. 2), and so all cRPI indents were included in the average calculation, instead of just those indents that overlapped with impact indents or the ROIs.

Figure 3.

Test for whether BMSi depended on location of indent. Except for the most distal and most proximal sites, the relative BMSi was the same on average among the sites of indentation. Each symbol represents a unique donor (*p<0.05 location 1 vs. other locations from Holm-Sidak’s test).

3.2. Anatomical differences in bone properties

Mean BMSi was lower for both arm bones than for the tibia mid-shaft (Table 1), whereas total indentation distance (TID) and average creep indentation distance (AvgCID) were higher for just the proximal humerus compared to the tibia mid-shaft (TMS). With respect to the left side, BMSi and average energy dissipated (AvgED) were significantly different between the distal radius (DR) and PH such that that PH had a lower BMSi and higher AvgED than DR (Table 1). There were no anatomical differences in indentation distance increase (IDI), average unloading slope (AvgUS), and average loading slope (AvgLS). Moreover, the anatomical differences in BMSi did not depend whether the left or right side was indented (Supplemental Fig. 3).

Table 1.

Anatomical differences in selected reference point microindentation properties as well as in local and regional microstructural properties as determined by µCT and ¹H NMR. Median (25%, 75% Quartiles).

		Tibia mid-shaft (TMS)		Distal Radius(DR)		Proximal Humerus 19		p-value
Property	Unit							TMS vs. DR	TMS vs. PH	DR vs. PH
OsteopProbe
BMSi	%	89.8	(84.4, 94.2)	79.9	(69.6, 87.6)	66.8	(62.3, 79.7)	0.0181	0.0003	0.0181
BioDent
IDI (1st-L)	µm	12.7	(11.8, 14.9)	12.1	(11.8, 13.9)	13.6	(12.2, 15.3)	ANOVA p=0.780
TID (1st-L)	µm	78	(75.4, 82.4)	80.7	(75.9, 95.1)	89	(82.7, 110.0)	NS	0.0417	NS
AvgCID (3rd-L)	µm	1.17	(1.12, 1.29)	1.26	(1.16, 1.33)	1.35	(1.28, 1.38)	NS	0.0417	NS
AvgED (3rd-L)	µJ	34.2	(33.2, 36.4)	33.6	(32.5, 37.7)	38.4	(37.3, 41.3)	NS	NS	0.0219
AvgUS (3rd-L)	N/µm	0.6	(0.56, 0.68)	0.61	(0.59, 0.67)	0.6	(0.58, 0.71)	ANOVA p=0.222
AvgLS (3rd-L)	N/µm	0.47	(0.45, 0.53)	0.49	(0.46, 0.51)	0.47	(0.45, 0.49)	ANOVA p=0.331
Microstructure (Local)
Ct.Po (box)	%	5.0	(4.0, 9.0)	16.0	(11.0, 26.0)	14.5	(8.5, 19.3)	0.0219	NS	NS
Ct.TMD (box)	mgHA/cm3	1051	(1030, 1078)	1072	(1061, 1085)	1043	(1027, 1057)	NS	NS	0.0024
Ct.avBMD (box)	mgHA/cm3	988	(977, 1017)	933	(802, 951)	911	(860, 958)	0.0472	0.0255	NS
Microstructure (Regional)
Ct.Th (manual contour)	mm	2.93	(1.46, 3.88)	1.28	(1.03, 1.57)	1.59	(1.01, 1.91)	0.0082	0.0082	NS
Pore water (NMR)	%	8.4	(5.9, 11.3)	18.1	(12.0, 22.7)	22.0	(17.7, 29.6)	0.0061	0.001	0.0262
Bound water (NMR)	%	18.6	(17.1, 19.6)	15.5	(13.7, 17.5)	15.6	(14.4, 17.6)	0.0468	0.0443	NS

To identify possible explanations for the indentation differences among bony sites, we examined the underlying intra-cortical material and microstructural properties of bone. The local apparent volumetric bone mineral density (avBMD) directly below the indents, the cortical thickness (Ct.Th), and regional bound water volume fraction were significantly lower in the arm bones than in the TMS (Table 1). Also, regional pore water volume fraction, a surrogate of regional porosity (i.e., not local periosteal porosity as determined by the µCT analysis), was significantly higher in PH than in DR (Table 1). Hence, the lower BMSi of PH compared to DR may be due to the significantly higher regional intra-cortical porosity for the PH than for DR. On the other hand, the higher AvgED for PH than for DR may be due to the lower local tissue mineral density (TMD) of PH compared to DR.

3.3. Correlations with microstructure

The local porosity explained greater than 48% of the variance in BMSi for the DR and PH (Fig. 4a). Higher porosity was associated with lower BMSi for both arm bones. The local TMD was negatively correlated to BMSi of the DR (Fig. 4b). The slopes of the linear regressions (BMSi vs. porosity or TMD) for DR and PH were not statistically significant (p=0.460 and 0.138, respectively), but the y-intercept terms were different (p<0.0001 and <0.0001, respectively) such that i) BMSi of DR was higher than BMSi of PH for a given local porosity and ii) local TMD of PH was lower than the TMD of DR for a given BMSi. There was also a significant correlation between BMSi and Ct.Th of the DR (Fig. 4c). BMSi of TMS did not depend on avBMD (Fig. 4d). Strikingly, AvgED was the only cRPI property to significantly correlate with µCT-derived properties, and unlike BMSi, these correlations occurred only in the tibia, not the arm bones (Table 2). Higher TMD and higher Ct.Th were associated with lower and higher AvgED, respectively.

Figure 4.

Relationships between BMSi and local µCT-derived properties. As Ct.Po increased, BMSi decreased for the arm bones (a). BMSi was also negatively correlated with TMD when excluding on outlier with very high TMD (b). BMSi was positively correlated with Ct.Th (c) and avBMD (d) for DR and for all bones combined. (Solid line: p<0.05 and Dashed line: p<0.1)

Table 2.

Pearson’s correlation coefficients between indentation resistance and microstructural properties for p-values less than 0.1 or 0.05.

Property	Unit	Tibia mid-shaft				Distal radius				Proximal humerus
		Ct.Po	TMD	avBMD	Ct.Th	Ct.Po	TMD	avBMD	Ct.Th	Ct.Po	TMD	avBMD	Ct.Th
BMSi	%	—	—	—	—	−0.90	−0.89a	0.88	0.77	−0.70	−0.60	0.63	0.66
		0.161	0.990	0.154	0.135	<0.001	<0.001	<0.001	0.008	0.009	0.064	0.039	0.063
IDI	µm	—	—	—	0.65	—	—	—	—	—	—	—	—
		0.667	0.370	0.975	0.068	0.937	0.844	0.866	0.684	0.884	0.731	0.900	0.996
TID	µm	—	—	—	—	—	—	—	—	—	—	0.55	0.63
		0.959	0.532	0.990	0.368	0.923	0.966	0.904	0.316	0.105	0.639	0.098	0.093
AvgCID	µm	—	—	—	0.62	—	—	—	—	—	—	—	—
		0.566	0.376	0.940	0.088	0.964	0.266	0.808	0.465	0.314	0.804	0.358	0.322
AvgED	µJ	—	−0.64	—	0.70	—	—	—	—	—	—	—	—
		0.152	0.018	0.600	0.008	0.785	0.769	0.909	0.295	0.560	0.482	0.676	0.303
AvgUS	N/µm	—	—	—	—	—	—	—	—	—	—	—	—
		0.446	1.000	0.335	0.435	0.548	0.406	0.229	0.968	0.997	0.490	0.998	1.000
AvgLS	N/µm	—	—	—	—	—	—	—	—	—	—	—	—
		0.150	0.934	0.119	0.103	0.999	0.215	0.882	0.149	0.993	0.849	0.917	0.754

^aOne donor was excluded from the correlation analysis because the TMD was 7 standard deviations from the mean (only for DR).

3.4. Correlations with material properties (Tibia mid-shaft only)

Since BMSi is the current measurement being assessed in vivo for human subjects, we tested to see if this microindentation resistance property of tissue was related to quasi-static mechanical properties of cortical bone from the TMS near the site of IMI. Even though the loading rates were vastly different between quasi-static flexural testing and dynamic impact microindentation, a positive correlation between BMSi and bending strength was statistically significant, but not so between BMSi and estimated toughness (Fig. 5). Not surprisingly, the coefficient of determination for bending strength and BMSi was not as high as those between bending strength and porosity of the beam (R²=96.0%, p<0.0001) or between bending strength and avBMD of the beam (R²=96.8%, p<0.0001).

Figure 5.

Linear regression between BMSi and paired mechanical properties of cortical bone at the apparent-level. BMSi was positively correlated with peak bending strength (a), yield strength (b), flexural modulus (c), but not with toughness (d). (gray symbols indicate properties were from the right, instead of the left tibia).

3. Discussion

Reference point microindentation – cyclic or impact – is currently the only technique capable of directly assessing the mechanical competence of bone in vivo. All other techniques from X-ray imaging and ultrasound to magnetic resonance imaging with finite element analysis provide surrogates of fracture resistance. While these surrogates related to bone mass, cortical structure, trabecular architecture, structural strength, and mineral density are significantly different between osteoporotic and otherwise healthy bone, they do not unequivocally predict fracture risk¹⁹. IMI may help with fracture risk assessment as it provides a unique tissue characteristic related to the mechanical resistance of cortical bone to high-rate indentation. In the present study, we find that the characteristic known as BMSi can directly correlate with material strength of cortical bone from elderly donors, an at risk population. Moreover, BMSi, unlike cRPI properties, is more sensitive to underlying intra-cortical microstructure, especially for bones with thin cortices and high cortical porosity throughout the cortex.

The mechanisms by which bone resists indentation remain to be fully elucidated for cRPI and IMI testing. However, there is evidence that certain cRPI (BioDent) measurements acquired at the TMS are different between osteoporotic fracture cases and non-fracture cases⁴ as well as between fracture cases (typical and sub-trochanteric) and age-and gender-matched controls²⁰. The IMI (OsteoProbe) measurement BMSi differs between type 2 diabetic and non-diabetic post-menopausal women²¹, negatively correlates with duration of type 2 diabetes²², changes from baseline following 7 weeks and 20 weeks of treatment with either denosumab (anti-RANKL) or teriparatide (hPTH(1–34))²³, differ between Norwegian and Spanish post-menopausal women but does not vary with age²⁴, correlate with subcutaneous fat in the lower limb²⁵, and are different between fracture and non-fracture cases with osteopenia and varies with age²⁶. Thus, there is promise that cRPI and IMI can ascertain aspects of bone tissue quality in a minimally invasive way to improve the prediction of fracture risk.

Nonetheless, there are still questions as to whether microindentation at the tibia mid-shaft is indicative of osteoporotic fractures, which typically occur at the hip, distal radius, proximal humerus, and spine. One population-based study to date found no difference in BMSi between women with a confirmed vertebral fracture and women without a fracture (both groups between 75 – 80 years), even though hip areal bone mineral density was significantly lower in the fracture group than in the controls²⁷. Although there were limitations to the study design including variable time between fracture event and the indentation tests as well as variance in BMSi among the 4 different operators of the OsteoProbe, the study points to the challenge of assessing tissue properties at the TMS to predict vertebral fracture, which also depends on the deterioration of the trabecular architecture. In our elderly cohort of 10 donors pooling the left and right sides with donor as a random effect, BMSi of TMS explained only 20.5% (p = 0.033) of the variance in the BMSi of the DR and was not related to the BMSi of the PH and (p = 0.330). This does not necessarily indicate that BMSi at the tibia is not predictive of fracture risk at the wrist or shoulder, but rather BMSi comparisons across groups should be made within a given anatomical site.

In two previous cadaveric studies involving both cRPI and IMI at the tibia mid-shaft (same quadrant as present study), BMSi was positively correlated with cortical TMD of the cross-section (r = 0.44, p = 0.055)⁶ and positively correlated with regional TMD (r = 0.43, p = 0.002, pooled left and right tibiae)¹². Scanning a cortical strip at a higher nominal resolution than the previous studies (10 µm compared to 82 µm and 30 µm) and confining the evaluation of TMD to the region below the indents, we found no significant correlation between BMSi and TMD of the tibia mid-shaft (Fig. 4). Interestingly, for the DR, there was a significant negative correlation between BMSi and TMD (Fig. 4). This suggests BMSi is not akin to tissue hardness, which typically has a positive correlation with degree of mineralization²⁸.

BMSi was previously found to weakly correlate with cross-sectional porosity (r = −0.37, p = 0.080)⁶ and regional porosity (r = −0.299, p = 0.037)¹². Here, in which local TMD and local Ct.Po did not correlate with one another (p = 0.646), the correlation between BMSi of TMS and local Ct.Po was also negative but not significant (p = 0.140). Local Ct.Po was inversely correlated with BMSi of DR and PH (Fig. 4) indicating that microstructure of the cortical bone influences the penetration depth of IMI.

In the aforementioned study by Karim et al.⁶, none of the cRPI properties (e.g., IDI, TID, AvgED) correlated with Ct.Po and TMD of the tibia cross-section (82 µm voxel size), while Abraham et al.¹² found significant correlations, albeit weak, between IDI and regional TMD (r = −0.390, p = 0.006) and between IDI and regional Ct.Po (r = 0.290, p = 0.043). In the present study analyzing bone from elderly donors, only AvgED significantly correlated with TMD and did so for only TMS (Table 2). AvgED is the energy dissipated (area under load-dwell-unload curve) averaged across the third to last cycle. Since the BioDent loads the bone in force control, AvgED typically increases as IDI and TID increase (r = 0.764, p = 0.008 and r = 0.664, p = 0.036), respectively) and so higher AvgED likely corresponds to lower indentation resistance (IDI did not correlate with unloading or loading stiffness, p > 0.12). Thus, the resistance to cyclic microindentation increased as TMD increased (i.e., positive association). Interestingly, AvgED and local TMD were both different between just DR and PH (Table 1) such that AvgED of DR with the higher TMD was lower than the AvgED of PH with the lower TMD. Broadly, cRPI appears to be i) less sensitive to underlying porosity than BMSi as none of the cRPI properties are different between TMS and DR even though local Ct.Po is significantly higher in DR and ii) more sensitive to TMD in which, combining all 3 bones, AvgED negatively correlates with TMD (r=−0.49, p=0.018) while BMSi does not correlate with TMD (combined: p=0.906).

Despite the given name of the measurement from IMI, the present work is the first to report a significant correlation between bone material strength index acquired at the tissue-level and material strength of human cortical bone at the apparent-level. Apparent bone mineral density, a product of porosity and degree of mineralization, primarily dictates such bending strength measurements. Given that BMSi did not correlate with local avBMD or Ct.Po of TMS and that the loading rates of bone tissue are vastly different between IMI and quasi-static three-point bending tests, the significant relationships are likely not causal (i.e., high BMSi does not necessarily confer high bending strength). Instead, donors with weak cortical bone from a material mechanics perspective also likely have low resistance to impact microindentation for other reasons unrelated to porosity and apparent bone mineral density.

In our previous work applying cRPI directly to mechanical specimens extracted from the cortex of human femurs, IDI was negatively correlated with bending strength (r = 0.50, p = 0.003) and estimated toughness (r = 0.49, p = 0.003)⁹. However, drying bone or removing the organic phase, both of which make cortical bone brittle, caused a reduction in IDI, not the expected increase if IDI is inversely related to toughness²⁹. Moreover, for rodent models in which bone brittleness is a dominant characteristic, IDI was found to be higher (type 1 diabetes³⁰, chronic kidney disease³¹, and osteogenesis imperfecta¹¹) or not different (advanced aging³² and disruption in crosslinks by lathyrism³³) compared to respective controls with normal bone toughness. Also, IDI was higher in a mouse model of high ductility (i.e., low brittleness due to hypomineralization)¹¹. Thus, the relationship between properties of microindentation resistance and mechanical behavior of cortical bone at the apparent level is context dependent in which different tissue characteristics (composition and organization) differentially affect indentation resistance. The present study suggests that the underlying microstructure could be another factor influencing indentation resistance of human cortical bone, especially for IMI.

A limitation of the present study was that it was not powered for multivariate analysis to determine whether a combination of microstructural features (e.g., porosity and cortical thickness) explained the differences in BMSi among the 3 anatomical sites. Nonetheless, BMSi is more sensitive to the differences in microstructure among 3 selected bones than any of the cRPI properties. While clinical studies are necessary to demonstrate the efficacy of measuring BMSi at the tibia to predict fracture sites at other sites, the present indicates that intraoperative measurements at the distal radius and proximal humerus may reflect changes in intra-cortical microstructure. In summary, while tissue resistance to cyclic microindentation at 10 N with respect to energy dissipation is positively related to tissue mineral density but not porosity, tissue resistance to impact microindentation at 40 N does appear to depend on the underlying porosity, at least for arm bones. The indentation characteristic BMSi can directly correlate with the apparent material strength of cortical bone from elderly donors. However, BMSi does not appear to be a measure of estimated toughness or the ability of the aged cortical bone tissue to dissipate energy, another major component of fracture resistance.

Supplementary Material

Supplemental Figures

Supplemental Figure 1: Micro-computed tomography analysis of machined cortical bone from TMS after the three-point bending test. To determine the distance between the surface and the center of mass (C_min), cross-sectional area (Area), and the moment of inertia (MOI) with respect to the axis of bending (I_min), evaluations were performed such that the entire region within the contour was segmented as bone including pores (a). Four of the 10 beam specimens had excessive trabecularization on the endosteal surface and were replaced with beams from the contralateral side. Note that the porosity can be higher near the endosteal surface (bottom left of panel a) than the periosteal surface (top left of panel a) such that regional pore water can be higher than local porosity.

Supplemental Figure 2: Relative cRPI properties at the eight neighboring indent locations across the different anatomical sites. None of the cRPI properties from BioDent varied significantly with indent location.

Supplemental Figure 3: Bone material strength index (BMSi) for the 3 anatomical sites comparing the left and right side. Top graph shows a Box plot representation with Tukey whiskers, and bottom graph shows the same data represented as a Mean Scatter plot. There was no significant difference in BMSi between the left or right sides of any of the bones that were indented (groups with different letters have significantly different BMSi between them).