The relationship between microstructure, stiffness and compressive fatigue life of equine subchondral bone

Journal of the Mechanical Behavior of Biomedical Materials 101 (2020) 103439

Contents lists available at ScienceDirect

Journal of the Mechanical Behavior of Biomedical Materials journal homepage: www.elsevier.com/locate/jmbbm

The relationship between microstructure, stiffness and compressive fatigue life of equine subchondral bone

T

Sandra Martiga,1, Peta L. Hitchensa, Peter V.S. Leeb, R. Chris Whittona,∗ a U-Vet Equine Centre, Melbourne Veterinary School, Faculty of Veterinary and Agricultural Sciences, The University of Melbourne, 250 Princes Highway, Werribee, VIC, 3030, Australia b Melbourne School of Engineering, Department of Biomedical Engineering, The University of Melbourne, Parkville, VIC, 3010, Australia

A R T I C LE I N FO

A B S T R A C T

Keywords: Subchondral bone Fatigue life Mechanical properties Micro-CT Equine Metacarpus

Subchondral bone injuries often precede articular cartilage damage in osteoarthritis and are common in thoroughbred racehorses due to the accumulation of fatigue damage from high speed racing and training. Thus, racehorses provide a model to investigate the role of subchondral bone in joint disease. We assessed the association of horse and racing related factors and micro-CT based micromorphology of three separate subchondral bone layers with the initial stiffness and compressive fatigue life of bone plugs. Furthermore, we investigated three different definitions of fatigue failure of subchondral bone during compressive fatigue testing. Initial stiffness was 2,362 ± 443 MPa (mean ± standard deviation). Median compressive fatigue life during cyclic loading to -78 MPa was 16,879 (range 210 to 57,064). Subchondral bone stiffness increased over a median of 24% (range 3%–42%) of fatigue life to a maximum of 3,614 ± 635 MPa. Compressive fatigue life was positively associated with bone volume fraction in the deeper layers of subchondral bone, maximal stiffness, and the number of cycles to maximal stiffness. Initial stiffness was positively associated with tissue mineral density in the deeper layers and bone volume fraction in the superficial layer. Most specimens with a fatigue life of less than 5,500 cycles fractured grossly before reaching 30% reduction of maximal stiffness. Cycles to 10% reduction of maximal stiffness correlated strongly with cycles to lowest recorded stiffness at gross fracture and thus is a valid alternative failure definition for compressive fatigue testing of subchondral bone. Our results show that subchondral bone sclerosis as a result of high speed exercise and measured as bone volume fraction is positively associated with compressive fatigue life and thus has a protective effect on subchondral bone. Further research is required to reconcile this finding with the common collocation of fatigue damage in sclerotic subchondral bone of racehorses.

1. Introduction Subchondral bone has been implicated in the development of osteoarthritis with changes in the subchondral bone often preceding and in some instances leading to articular cartilage damage (Lajeunesse et al., 2003; Radin et al., 1973; Li et al., 2013). Thoroughbred racehorses commonly develop subchondral bone fatigue injuries in the palmar aspect of the metacarpo (metatarso)-phalangeal joints (Barr

et al., 2009; Pinchbeck et al., 2013; Norrdin and Stover, 2006; Turley et al., 2014). In this location, tendon wrapping generates extreme cyclical loads between the sesamoid bones and the metacarpal (metatarsal) condyles during high speed training and racing (Harrison et al., 2010). Hence racehorses provide a model to investigate the role of subchondral bone in joint disease. In young growing racehorses the subchondral bone plate thickens due to adaptive modelling within two months of the start of training

Abbreviations: aclporestra, area of closed pores in cross-section (average of 100 images); BV/TV, bone volume fraction (bone volume divided by tissue volume); BS/ BV, specific bone surface (bone surface divided by bone volume); BS/TV, bone surface density (bone surface divided by tissue volume); clporestra, number of closed pores in cross-section (average of 100 images); Ei, initial Young's modulus (Young's modulus prior to fatigue testing); Em, maximal Young's modulus during fatigue testing; hBV/TV, volume fraction of highly mineralised tissue; hTMD, tissue mineral density of highly mineralised tissue; lBV/TV, volume fraction of less mineralised tissue; lTMD, tissue mineral density of less mineralised tissue; SCBI, subchondral bone injury; VOI, volume of interest ∗ Corresponding author. E-mail addresses: [email protected] (S. Martig), [email protected] (P.L. Hitchens), [email protected] (P.V.S. Lee), [email protected] (R.C. Whitton). 1 present address: Centre for Animal Referral and Emergency, 5 Hood Street, Collingwood VIC, 3066, Australia. https://doi.org/10.1016/j.jmbbm.2019.103439 Received 13 May 2019; Received in revised form 16 September 2019; Accepted 16 September 2019 Available online 17 September 2019 1751-6161/ © 2019 Published by Elsevier Ltd.

Journal of the Mechanical Behavior of Biomedical Materials 101 (2020) 103439

S. Martig, et al.

2. Material and methods

(Muir et al., 2008; Rubio-Martinez et al., 2008a; Riggs and Boyde, 1999; Boyde and Firth, 2005). This physiological response presumably enables the bone to withstand the cyclic loading of training and racing, an assumption further corroborated by the positive correlation between bone volume fraction and compressive fatigue life in cancellous bone (Mazurkiewicz and Topolinski, 2017; Rapillard et al., 2006; Topolinski et al., 2011; Fatihhi et al., 2015). Similar studies of fatigue properties in subchondral bone are lacking, and associations between structural and monotonic mechanical properties of subchondral bone are inconsistent. For example, higher subchondral bone volume fraction leads to increased stiffness at relatively low strain rates whereas this association does not persist at high strain rates (Rubio-Martinez et al., 2008b; Malekipour et al., 2018; Martig et al., 2013). Furthermore, higher subchondral bone volume fraction is consistently associated with fatigue damage in mature racehorses prompting speculation that this thickening of the subchondral bone plate contributes to injury (RubioMartinez et al., 2008a; Muir et al., 2006; Whitton et al., 2018). However, it is difficult to reconcile both a protective and a detrimental effect of subchondral bone modelling on joint health. Improved knowledge of the effect of bone morphology on subchondral bone compressive fatigue life and stiffness is required to better understand the development of subchondral bone injuries. In vitro compressive fatigue testing is routinely performed in cancellous bone. In order to define the point of failure of each specimen a percentage of stiffness reduction compared to a reference cycle at the start of fatigue loading is often used (Rapillard et al., 2006; Dendorfer et al., 2008; Bowman et al., 1998; Haddock et al., 2004). This is based on the observation that cancellous bone stiffness decreases during fatigue testing and any increase in stiffness is assumed to be artefactual or due to compaction of grossly fractured trabeculae, thus after failure (Bowman et al., 1998; Moore and Gibson, 2003; Dendorfer et al., 2006). We previously performed compressive fatigue testing of subchondral bone at different loads and found that most of our subchondral bone specimens showed an increase of stiffness during the first half of fatigue life, i.e. during several hundreds to thousands of loading cycles (Martig et al., 2013). Hence, stiffness determined from a predetermined cycle at the start of fatigue testing proved unsuitable as a reference to define failure. Consequently, we defined the cycle of failure as the cycle with the lowest recorded stiffness in cyclic compressive loading terminated when the testing machine actuator displacement reached a predetermined machine safety limit that corresponded with gross specimen fracture (Martig et al., 2013). This method is repeatable in our hands but requires estimation of the expected actuator displacement at gross fracture based on trial specimens, as actuator displacement at the point of gross specimen fracture varies with strain rate or load. A different definition of failure during compressive fatigue testing of subchondral bone is therefore desirable. We performed a cross-sectional study investigating mechanical properties of subchondral bone from the palmar aspect of the distal metacarpal condyle in thoroughbred racehorses in training with the following two aims. First, to determine how horse and racing career related indices and micro-CT derived morphological parameters are correlated with stiffness and compressive fatigue life. We hypothesised that fatigue life and initial stiffness would increase with increasing bone volume fraction. Second, to compare the number of cycles to fatigue failure during compressive cyclic loading at a) 10% and b) 30% reduction of maximal stiffness and c) the lowest recorded stiffness when the actuator displacement reaches a predetermined machine safety limit that corresponds with gross specimen fracture. We hypothesised that there would be good correlation between fatigue failure defined as (a) the cycle where 10% reduction of maximal stiffness is reached, (b) the cycle where 30% reduction of maximal stiffness is reached and (c) the cycle with the lowest recorded stiffness in a test run, when the testing machine actuator displacement reached a predetermined machine safety limit that corresponded with gross specimen fracture.

2.1. Horses Metacarpal condyles were collected within 24 h of death from a convenience sample of Thoroughbred racehorses that died or were euthanised on racetracks in Victoria, Australia, during racing or training between March 2010 and November 2013. These horses underwent compulsory post mortem examination at the pathology department of The University of Melbourne, Faculty of Veterinary and Agricultural Sciences, and specimen sampling was in accordance with guidelines of The University of Melbourne animal ethics committee (Ethics ID #14001). Horses were included if they were at least two years old, did not meet inclusion criteria for other studies which required the metacarpi, staff were available to collect samples and the trainer provided information about the number of weeks the horse was in training since the last period of rest. Twenty-four horses were included in the study. For each horse the time in training since the most recent period of rest (in weeks), the number of life time race starts and the cause of death (euthanasia due to musculoskeletal fatigue injury or other) were recorded. Subchondral bone injury (SCBI) was graded as 0 (no lesions), 1 (subchondral bone discoloration, normal overlying cartilage), 2 (subchondral bone discoloration and overlying cartilage lesion), and 3 (subchondral bone and overlying cartilage defect). Storage time in months was calculated based on collection date and mechanical testing date. One forelimb per horse was chosen with the use of a random number table (White et al., 1977). In horses with trauma to a metacarpophalangeal joint, the contralateral metacarpophalangeal joint was used (n = 3, all with trauma to the left metacarpophalangeal joint). 2.2. Specimen preparation Bones were individually wrapped in saline soaked gauze, placed in ziplock bags and stored in plastic containers at -20° Celsius after collection and between steps of preparation, micro-CT and mechanical testing. Cylindrical bone specimens from the palmar aspect of the medial distal metacarpal condyle were prepared as previously described (Martig et al., 2013). Medial metacarpal condyles are more commonly affected by higher grades of SCBI than lateral condyles, and due to requirements for other experiments more medial than lateral condyles were available (Pinchbeck et al., 2013). The palmar aspect of the metacarpal condyle was sectioned in a proximo 55° palmar to distodorsal oblique plane (Fig. 1). The area of interest for the cylindrical bone specimen was located at the centre of the medial condyle slab. The long axis of the cylinder was approximately orthogonal to the cut surface of the slab corresponding to the loading axis of the articulation between the proximal sesamoid bone and the metacarpal condyle during the stance phase at high speed gallop. The articular cartilage was removed with a scalpel blade. A diamond coated core drill (#102075, Starlite Industries Inc, Rosemont, PA, USA) was used to drill the bone core at room temperature under constant irrigation with tap water. The resulting cylinders had a diameter of approximately 6.3 mm and were cut to a length of approximately 7.4 mm with a diamond coated wafer blade on a low speed precision saw (IsoMet, Buehler, Lake Bluff, Illinois, USA) using tap water for cooling. The size of the specimen diameter allowed for a grossly flat articular surface in the majority of specimens. The ends of the cylinders were wet ground manually in a custom jig until planoparallel using electrocoated silicon carbide abrasive paper (P600, KMCA, Bunnings, Hawthorn, VIC, Australia). 2.3. Micro-computed tomography Specimens were thawed to room temperature and imaged with an 2

Journal of the Mechanical Behavior of Biomedical Materials 101 (2020) 103439

S. Martig, et al.

mineralised bone. The average value of the 100 distal most slices in each VOI was calculated for both the number and area of closed pores. The 100 distal most slices with maximum diameter were used in the distal VOI. The original greyscale images were then segmented again, this time using the two-level Otsu method in CTAn, which provided two segmentation levels (Otsu, 1979). The level 1 Otsu grey value segments mineralised from non-mineralised tissue and was 58 ± 4.3 (n = 68). For consistency, level 1 Otsu threshold was replaced with the above described manually determined lower grey threshold. The level 2 Otsu threshold separates mineralised tissue into highly mineralised (level 2 Otsu threshold to 255) and less mineralised tissue (lower grey threshold to [level 2 Otsu threshold–1]). The level 2 Otsu grey value was 82 ± 3.8 (n = 69). The tissue mineral density and volume fraction of the highly mineralised tissue (hTMD and hBV/TV, respectively) and the less mineralised tissue (lTMD and lBV/TV, respectively) were determined as above. 2.4. Mechanical testing The mechanical testing order was determined by a random draw of specimen numbers. Mechanical testing was performed on a hydraulic material testing machine (Instron 8874, Instron, Norwood, MA, USA) at room temperature (18 °C–21 °C) with the specimen submerged in Cabuffered 0.9% saline (0.154 M NaCl, 1.381 mM CaCl2) in a setup as previously described (Martig et al., 2013). Prior to mechanical testing, specimens were thawed to room temperature for 30 min in Ca-buffered 0.9% saline. The trabecular end of the thawed specimen was potted into the rigid stainless steel lower platen by placing it in the 2 mm deep opening of the plate and filling the space around it with polymethylmethacrylate (PMMA) premixed with a powder to liquid ratio of 2:1 (Giltspur Scientific Ltd, Ballyclare, Northern Ireland). After potting, specimens were covered with Ca-buffered 0.9% saline and cured for 30 min before mounting onto the material testing machine. The articular surface was in contact with the solid, non-porous stainless steel upper platen without end-constraint or lubrication. The upper platen was fastened in series with the 10 kN load cell (resolution ± 0.001; Dynacell, Instron, Norwood, MA, USA) on the actuator of the material testing machine and the recordings of the load cell were used for stress calculations. Displacement of the upper platen was recorded by the machine's inbuilt linear variable displacement transducer. Engineering strain was calculated as the displacement divided by the free gauge length of the specimen. Half of the PMMA embedded specimen length (i.e. 1 mm) was added to the free gauge length for strain calculations to account for the stiffening effect of embedding (Keaveny et al., 1997). Engineering stress was calculated as the force applied by the load cell divided by the specimen surface area. Bone stiffness was measured in the form of Young's modulus, calculated as engineering stress divided by engineering strain. Young's modulus was automatically calculated by the software (Wavematrix version 1.4.268.0, Instron, Norwood, MA, USA) using a least squares fit and 25% of the compression part of the loading curve as lower and 75% as upper limits for calculations. Data acquisition frequency was 500 Hz with the resampling filter set on automatic. Recorded data were saved for every cycle up to 100 cycles, then every 10th cycle up 1000 cycles followed by every 100th cycle until the test ended when the set actuator displacement safety limited was tripped. Data were also saved for the last 100 cycles before test end. The compression-compression testing sequence using a sinusoidal waveform consisted of preconditioning under load control, determination of initial Young's Modulus under displacement control, and fatigue testing under load control as previously described but with the following modifications (Martig et al., 2013). The loading frequency was 2.4 Hz for preconditioning and fatigue testing to more closely mimic the physiological stride frequency at racing speed (Witte et al., 2006; Seder and Vickery, 2003). In order to eliminate any strain hardening the upper platen was removed from the specimen for complete unloading of at least 2 min prior to initial Young's modulus determination and prior

Fig. 1. Volume rendered CT image of the palmar aspect of equine distal metacarpal condyles demonstrating location of bone specimens used in this study. Line (a) indicates the long axis of the metacarpal diaphysis. Bone slabs were cut along line (b), which intersects with line (a) at the estimated centre of rotation of the condyles at an angle alpha of 55°.

ex vivo micro-CT scanner (SkyScan 1172, Bruker microCT N.V., Kontich, Belgium) at a pixel size of 4.87 μm as described previously (Martig et al., 2018). Briefly, specimens were scanned in air in a high humidity environment ensuring hydration of the specimens was maintained throughout the 38 min of scan duration. Temperature at the specimen table reached approximately 27° Celsius during scanning and specimens were prepared for storage and transferred onto ice immediately afterwards and refrozen as soon as practical. Images of calcium hydroxyapatite phantoms (Bruker-microCT, Kontich, Belgium) were used to calibrate the software (CTan 1.13.15.1+, Bruker-microCT, Kontich, Belgium) for bone mineral density calculations. Micro-CT images were reconstructed, aligned and analysed using proprietary software (image reconstruction: Nrecon version 1.6.6.0, Bruker-microCT, Kontich, Belgium; image alignment: DataViewer 64 bit, version 1.4.4.4, Skyscan, Kontich, Belgium; image analysis: CTan 1.13.15.1+, Bruker-microCT, Kontich, Belgium) as previously described (Martig et al., 2018). Briefly, each specimen was divided into three contiguous volumes of interest (VOI) of equal diameter (5.3 mm) and length (1.7 mm). The distal VOI was chosen to include the distal most image that contained articular surface. The VOI diameter of these distal most images was adjusted to accommodate the curvature of the articular surface resulting in slight variation of volume of the distal VOI (mean 34.8 μm3) while all middle and proximal VOIs had a volume of 37.8 μm3. Bone volume fraction (BV/TV, in %), specific bone surface (BS/BV in mm−1), bone surface density (BS/TV in mm−1) and tissue mineral density (TMD, in g/cm3) were determined for each VOI after initial segmentation. Images were segmented using a manual global thresholding method to best match the segmented images with the greyscale images. Thresholding values were grey values derived from CTAn. The lower grey threshold separated non-mineralised from mineralised tissue and was determined for each VOI individually (mean ± standard deviation of all VOIs 54 ± 4.4, n = 69). The software required definition of an upper threshold, which was set to a grey value of 255 for all VOIs. The number of closed pores and the area of closed pores (in mm2) in cross-section were determined as 2D parameters. A closed pore on a cross-sectional (2D) image is an area of background pixel values (i.e. non-mineralised marrow spaces and vascular canals) that is completely enclosed by 3

Journal of the Mechanical Behavior of Biomedical Materials 101 (2020) 103439

S. Martig, et al.

and Kurtosis tests for normality and the Breusch-Pagan / Cook-Weisberg test for heteroscedasticity. Statistical analyses were performed using Stata/IC 15.1 (Statacorp, College Station, TX, USA) and P < 0.05 was considered significant. Post hoc power analyses were performed using G*Power 3.1.9.2 (Faul et al., 2009). Within the test family of F tests the following statistical test was chosen: linear multiple regression: fixed model, R2 increase. Effect size (f2 ) was calculated based on the correlation coefficient (R2) of the linear regression analysis. The alpha error probability was set at 0.05. For calculations of required sample size the type of power analysis was a priori, using the same f2 as for the post hoc calculations with power set at 0.08.

to fatigue testing. Preconditioning and initial Young's modulus determination were performed as previously described (Martig et al., 2013). Briefly, preconditioning included 50 cycles under load control from a small preload (between -20 to -50 N, corresponding to 0.6–1.4 MPa) to a compressive stress of 20 MPa. Initial Young's Modulus was determined as the mean Young's modulus of the last six of ten compression cycles from the above preload to a maximal compressive engineering strain of approximately 1.0% at a strain rate of 0.01s−1, loaded under displacement control. Fatigue testing was performed under load control from the above preload to a maximum compressive load of 78 MPa for all specimens. The maximal load of 78 MPa was chosen based on the fatigue life curve of similar specimens with the expected specimen fatigue lives to be in the tens of thousands of cycles, which was deemed acceptable regarding test time requirements while avoiding multiple specimens failing after a few hundred cycles (Martig et al., 2013). Cyclic loading was stopped when the actuator displacement from the preload position at the start of the fatigue test exceeded the predetermined safety limit of 1.6 mm. All specimens were grossly fractured at this displacement. The cycle of failure for the purpose of the fatigue test was defined as the cycle with the lowest recorded Young's modulus reading (Martig et al., 2013). The maximum recorded Young's modulus and the number of cycles to the maximum recorded Young's modulus were determined as well as the number of cycles to reach 10% and 30% reduction of the maximal recorded Young's modulus, respectively. One specimen was accidentally overloaded during set up for mechanical testing and was excluded from all analyses. Fatigue testing of one specimen was accidentally terminated prior to failure of the specimen and hence only initial Young's Modulus was available for analysis. One specimen was unloaded for less than 2 min prior to initial Young's modulus testing. Analyses were performed with and without this specimen. None of the explanatory variables experienced a change in effect direction or significance, thus we retained the specimen in all analyses. Three specimens had a subjectively more pronounced gross curvature of the articular surface than the remainder. Again, analyses were performed with and without these specimens. None of the explanatory variables experienced a change in effect direction or significance, thus we retained these specimens in all analyses.

3. Results 3.1. Horses Of the 23 horses remaining for analysis nine were castrated males and 13 were females. Archival information of one horse was insufficient to determine sex and the number of race starts. Six horses were 2-yearolds, ten were 3-year-olds and seven horses were 4-year-old or older. Seven horses were euthanised due to musculoskeletal stress injury [humerus fracture (n = 2), metacarpal diaphyseal fracture (unilateral n = 1, bilateral n = 1) and one each with a fracture of proximal sesamoid bone, bilateral carpal bones, and metatarsal condyle]. Causes of death for the remaining horses were sudden death (n = 5), exercise induced pulmonary haemorrhage (EIPH) (n = 4); combination of EIPH and radius fracture after collision with railing (n = 1); fall at a hurdle (n = 1); annular ligament avulsion during fast work (n = 1), anaphylactic reaction to injection (n = 2), colic (n = 1) and drowning (n = 1). Seven horses had SCBI affecting the specimen (grade 1, n = 4; grade 2, n = 2; grade 3, n = 1). Due to the small number of horses with grade 2 and grade 3 SCBI, SCBI was categorised as absent (n = 16) or present (n = 7) for statistical analyses. Horses were in training since their last period of rest for (mean ± standard deviation, 14 ± 7 weeks . The number of race starts was unknown for one horse. Fourteen of 22 horses started at least once in a race and had 11 ± 9 race starts. The proportion of raced and unraced horses reflects our training and racing related post mortem population. Bones were stored frozen for (mean ± standard deviation, range) 50 ± 11, 27–71 months.

2.5. Statistical analyses 3.2. Stiffness and fatigue life Descriptive statistics were calculated for all micro-CT parameters, initial Young's modulus, maximal Young's modulus, the number of cycles to maximum Young's modulus, the number of cycles to failure, the number of cycles to 10% reduction of maximal Young's modulus and the number of cycles to 30% reduction of maximal Young's modulus. Principal component analysis (PCA) of the micro-CT parameters was performed to reduce the number of parameters explaining the mechanical properties and yielded similar principal components as previously found for subchondral bone of the lateral metacarpal condyle (Supplementary Fig. 1). (Martig et al., 2018) Thus the same parameters, i.e. BV/TV, hBV/TV, TMD and number of closed pores, were chosen as representative micro-CT parameters for further analysis. Relationships between continuous explanatory and outcome variables were initially assessed based on scatter plots. Univariable linear regression was performed to investigate the association between the outcome variables (initial Young's modulus and the number of cycles to failure) and each explanatory variable (storage time, limb, age, sex, number of starts, weeks in training since last period of rest, subchondral bone injury, cause of death and the micro-CT derived parameters BV/TV, hBV/TV, TMD and number and area of closed pores for the distal, middle and proximal VOI). For the outcome variable number of cycles to failure we also assessed associations with maximal Young's modulus, the number of cycles to maximal Young's modulus and the additional test ends (number of cycles to 10% and 30% reduction in maximal Young's modulus, respectively). Model residuals were assessed using Skewness

Descriptive statistics for the micro-CT parameters are presented in Table 1 and for the mechanical properties in Table 2. Maximal Young's modulus was reached after a median of 24% (range, 3%–42%) of fatigue life. The most common pattern of change in Young's modulus during fatigue testing consisted of an initial rapid then slower increase in Young's modulus over a variable number of loading cycles, merging into a relatively stable phase of slowly decreasing Young's modulus, followed by a rapid final decrease in Young's modulus culminating in gross fracture. This pattern was observed in 20 of 22 specimens. The two remaining specimens had the shortest and 4th shortest fatigue life. Both specimens had an early increase, then rapid drop of stiffness. This was followed by a plateau and final rapid decrease in stiffness in one specimen. In the other specimen, Young's modulus increased again slowly, reached a maximum higher than the first maximum Young's modulus followed by a slowly decreasing plateau and steep final decrease of stiffness. The second, higher Young's modulus peak was used as maximum Young's modulus. However, gross failure may have occurred during the first rapid Young's modulus decrease despite it being less than 10% of the first peak of Young's modulus. Thus, analyses investigating factors affecting fatigue life were performed with and without this specimen. There were no changes in significance. Changes in effect direction were observed in two non-significant variables (3year-old and volume fraction of highly mineralised bone in the middle VOI) with little change in the confidence intervals of the effect. 4

Journal of the Mechanical Behavior of Biomedical Materials 101 (2020) 103439

VOI: volume of interest (5.3 mm in diameter and 1.7 mm in height); BV/TV: bone volume fraction; lBV/TV: volume fraction of less mineralised bone; hBV/TV: volume fraction of highly mineralised bone; TMD: tissue mineral density of all mineralised tissue; lTMD: tissue mineral density of less mineralised tissue; hTMD: tissue mineral density of highly mineralised tissue; clporestra: number of closed pores in cross-section (average of 100 slices); aclporestra: area of closed pores in cross-section (average of 100 slices); BS/TV: bone surface density; BS/BV: specific bone surface.

4.53 ± 1.19 6.01 ± 0.95 7.06 ± 1.23 4.40 ± 1.07 5.66 ± 0.80 6.35 ± 0.80 0.45 ± 0.35 0.92 ± 0.36 1.53 ± 0.60 97.3 ± 1.6 94.5 ± 2.0 90.4 ± 3.7 distal (incl. articular surface) middle proximal

54.6 ± 7.7 43.0 ± 3.2 37.8 ± 4.7

42.7 ± 6.4 51.4 ± 2.0 52.6 ± 1.7

0.92 ± 0.02 1.03 ± 0.03 1.07 ± 0.02

0.83 ± 0.02 0.92 ± 0.03 0.95 ± 0.02

1.03 ± 0.03 1.11 ± 0.03 1.16 ± 0.02

316 ± 99 411 ± 94 380 ± 87

BS/TV (mm−1) aclporestra (mm2) clporestra hTMD (g/cm3) lTMD (g/cm3) TMD (g/cm3) hBV/TV (%) lBV/TV (%) BV/TV (%) VOI

Table 1 Mean and standard deviation of micro-CT parameters of a subchondral bone samples from the palmar aspect of the medial metacarpal condyle of 23 thoroughbred racehorses in training.

BS/BV (mm−1)

S. Martig, et al.

Therefore, this specimen was retained for all analyses. 3.3. Results of statistical analyses Table 3 provides the results of the linear regression analyses assessing the association between the outcome variables (initial Young's modulus and cycles to failure) and the various explanatory variables. Initial Young's modulus was positively associated with bone volume fraction in the distal VOI (adjusted R2 = 0.24), volume fraction of highly mineralised bone in the proximal VOI (adjusted R2 = 0.21), and tissue mineral density in the middle (adjusted R2 = 0.34) and proximal VOIs (adjusted R2 = 0.30) (Fig. 2). Volume fraction of highly mineralised bone in the distal VOI was negatively associated with initial Young's modulus (adjusted R2 = 0.21) (Fig. 2). Bone volume fraction in the middle and proximal VOIs were positively correlated with the number of cycles to failure (adjusted R2 middle VOI = 0.28; proximal VOI = 0.21) (Fig. 2). Furthermore, the number of cycles to failure were also positively correlated with maximal Young's modulus (adjusted R2 = 0.50) and the number of cycles to maximal Young's modulus (adjusted R2 = 0.74). 3.4. Failure definition The cycle of lowest recorded Young's modulus corresponded with the cycle when the actuator displacement safety limit was tripped in 18 of 22 specimens. In one of the remaining specimens the lowest recorded Young's modulus was followed by an increase in Young's modulus prior to the safety limit being tripped. In the remaining three specimens the cycle with the lowest recorded Young's modulus was unlikely the cycle of specimen failure. In two of them, Young's modulus decreased towards the cycle during which the actuator displacement safety limit was reached but remained higher than Young's modulus at the start of fatigue testing. The cycle when the actuator displacement safety limit was tripped was used as cycle of failure and corresponded to the cycle with the lowest recorded Young's modulus after maximal Young's modulus. In the remaining specimen Young's modulus decreased rapidly after the plateau phase followed by a stepwise increase and another decreased to the cycle where the actuator displacement safety limit was tripped, which corresponded to the lowest recorded Young's modulus of the fatigue test. We assumed that specimen failure occurred during the initial fast decrease of Young's modulus and the cycle with the lowest recorded Young's modulus reading before the stepwise increase was used as cycle of failure. Five of the six specimens with less than 5500 cycles to failure fractured grossly before reaching 30% reduction of maximal Young's modulus (Fig. 3). Hence, only cycles to 10% reduction of maximal Young's modulus were compared to cycles to failure as possible alternative definition of fatigue failure. The number of cycles to 10% reduction of maximal modulus could be determined unequivocally for all specimens, taking into consideration that storing recorded data only at predefined intervals during fatigue testing provided an approximation of the exact cycle of maximal Young's modulus and 10% reduction of maximal Young's modulus, respectively. Cycles to 10% reduction of maximal Young's modulus (Nm-10) explained 99.7% of the variation of cycles to failure (Nf), described by the following linear regression equation: Nf = 516.7 + 1.1 x Nm-10, P < 0.001 (Fig. 4). In comparison, cycles to maximal Young's modulus, while also strongly positively associated with cycles to failure (Table 3), only explained 74% of the variation of cycles to failure. 4. Discussion We observed a complex relationship between subchondral bone microstructure and mechanical properties. Higher bone volume fraction in the distal VOI was associated with higher whole specimen stiffness, and higher bone volume fraction in the middle and proximal VOIs with 5

Journal of the Mechanical Behavior of Biomedical Materials 101 (2020) 103439

S. Martig, et al.

Table 2 Mechanical properties of subchondral bone from the palmar aspect of the medial metacarpal condyle of Thoroughbred racehorses in training.

Ei [MPa] (n = 23) Emax [MPa] (n = 22) NEmax (n = 22) Nf(Emin) (n = 22) Nf(Emax-10%) (n = 22) Nf(Emax-30%) (n = 17)

a

mean ± standard deviation

median

range

variance

2,362 ± 443 3,614 ± 635 5,722 ± 5,630 22,805 ± 18,708 20,003 ± 16,766 27,918 ± 16,889

2,472 3,704 3,400 16,879 14,150 23,100

1,412 to 3,266 1,950 to 4,547 7 to 20,5000 210 to 57,064 170 to 51,400 4,300 to 56,600

2.0 × 106 4.0 × 106 3.2 × 107 3.5 × 108 2.8 × 108 2.9 × 108

Ei: Young's modulus prior to fatigue testing; Emax: highest recorded Young's modulus during compressive fatigue testing; NEmax: cycles to Emax; Nf(Emin): cycles to failure (lowest recorded Young's modulus); Nf(Emax-10%): cycles to 10% reduction of Emax; Nf(Emax-30%): cycles to 30% reduction of Emax. a five specimens, all with Nf(Emin) < 5500, fractured grossly before reaching 30% reduction of Emax. Table 3 Results of univariable linear regression analyses of factors influencing Young's modulus and compressive fatigue life of subchondral bone from the palmar aspect of the medial metacarpal condyle of thoroughbred racehorses in training. explanatory variable

Ei (MPa) coef (95% CI) (n = 23)

p

Nf

storage time (months) right limb age 2-year-old 3-year-old 4-year-old and older sex female castrated male weeks in training starts SCBI cause of death bone volume fraction (%) distal middle proximal hBV/TV (%) distal middle proximal tissue mineral density (g/cm3) distal middle proximal number of closed pores distal middle proximal Ei (MPa) Emax (MPa) NEmax

1.8 (-17 to 21) −185 (-580 to 209)

0.8 0.3

−59 (-806 to 778) −10,454 (-27,493 to 6,585)

0.9 0.2

reference −97 (-590 to 396) −205 (-736 to 326)

0.7 0.4

reference −446 (-22,626 to 21,734) −7,344 (-31,055 to 16,367)

> 0.9 0.5

reference 2 (-403 to 406) −18 (-46 to 10) −15 (-37 to 7) (n = 22) −339 (-739 to 60) −32 (-445 to 381)

> 0.9 0.2 0.2 0.09 0.9

reference 10,015 (-7,082 to 27,112) 293 (-933 to 1,519) −595 (-1,573 to 384) (n = 21) −11,345 (-28,867 to 6,177) −3,520 (-21,750 to 14,710)

0.2 0.6 0.2 0.2 0.7

145 (39–250) 29 (-71 to 129) −29 (-82 to 25)

0.01 0.6 0.3

3,720 (-1,367 to 8,806) 5,197 (1,649 to 8,746) 2,533 (471–4,594)

0.1 0.006 0.02

−35 (-62 to -7) 55 (-43 to 154) 128 (25–232)

0.02 0.3 0.02

−780 (-2,120 to 560) 140 (-4,183 to 4,463) −241 (-5,357 to 4,875)

0.2 0.9 0.9

8,649 (-3,662 to 20,960) 9,520 (3,831 to 15,208) 10,331 (3,697 to 16,966)

0.2 0.002 0.004

−278,389 (-821,027 to 264,249) 156,198 (-137,750 to 450,146) 197,794 (-134,169 to 529,758)

0.3 0.3 0.2

−2 (-4 to 2) 0.4 (-1.8 to 2.5) −0.2 (-2.5 to 2.1) -

0.07 0.7 0.9

−34 (-124 to 55) 39 (-51 to 128) 28 (-74 to 130) 17 (-0.4 to 35) 21 (12–31) 3 (2–4)

0.4 0.4 0.6 0.06 < 0.001 < 0.001

(Emin)

coef (95% CI) (n = 22)

p

coef: coefficient; 95% CI: 95% confidence interval; Ei: Young's modulus (MPa) prior to fatigue testing; Emax: highest recorded Young's modulus (MPa) during fatigue testing; Nf(Emin): number of cycles to fracture (lowest recorded Young's modulus); SCBI: subchondral bone injury; distal: volume of interest (VOI, 5.3 mm in diameter, 1.7 mm high) including the articular surface; middle: VOI immediately adjacent to the distal VOI and of similar size; proximal: VOI immediately adjacent to the middle VOI and of similar size; hBV/TV: volume fraction of highly mineralised bone. NEmax: number cycles to Emax.

assessed, was not associated with fatigue life. The positive association of bone volume fraction with fatigue life is consistent with correlations reported in trabecular bone, though weaker (R2 for trabecular bone: 0.70 to 0.78) (Mazurkiewicz and Topolinski, 2017; Rapillard et al., 2006; Topolinski et al., 2011; Fatihhi et al., 2015). Due to the gradients in both bone volume fraction and tissue mineral density we previously identified in similar subchondral bone specimens, we chose to assess micromorphological parameters in separate VOIs rather than for the entire specimen (Martig et al., 2018). Whole specimen assessment of micromorphology may be more suitable for investigating whole specimen mechanical properties but ignores the complexity of subchondral bone structural and material properties. The actual differences in bone volume fraction between and within VOIs were relatively small. Further research is required to investigate whether the association between bone volume fraction and mechanical properties is causative and

longer compressive fatigue life. This corroborates our hypothesis that specimens with a thicker subchondral bone plate have a longer fatigue life but not that they are also stiffer initially. In fact, individual morphological parameters were associated with either initial stiffness or fatigue life but not with both mechanical properties. Additionally, initial specimen stiffness was not associated with compressive fatigue life whereas maximal stiffness reached during cyclic loading and the number of cycles to reach maximal stiffness were both positively associated with compressive fatigue life.

4.1. Factors affecting fatigue life Bone volume fraction in the middle and proximal VOIs were the only micromorphological explanatory parameters associated with fatigue life whereas tissue mineral density, another key bone variable 6

Journal of the Mechanical Behavior of Biomedical Materials 101 (2020) 103439

S. Martig, et al.

Fig. 2. Scatter plots of significant associations between mechanical and micro-CT parameters of equine subchondral bone specimens. (A) The association between cycles to fracture [Nf(Emin)] and bone volume fraction [BV/TV] is significant in the middle [Nf(Emin) = –468,626 + 5,197 * BV/TV, adjusted R2 = 0.28, P = 0.006] and proximal VOI [Nf(Emin) = –206,669 + 2,533, adjusted R2 = 0.21, P = 0.02] but not in the distal VOI [P = 0.1]. (B) The association between initial Young's modulus [Ei] and BV/TV was significant in the distal [Ei = –11,710 + 145 * BV/TV, adjusted R2 = 0.24, P = 0.01] but not in the middle (P = 0.6) and proximal VOIs (P = 0.3). (C) The association between Ei and volume fraction of highly mineralised bone [hBV/TV] was significant in the distal [Ei = 3,849 – 35 * hBV/TV, adjusted R2 = 0.21, P = 0.02] and proximal VOIs [Ei = – 4.391 + 128 * hBV/TV, adjusted R2 = 0.21, P = 0.02] but not in the middle VOI [P = 0.3). (D) The association between Ei and tissue mineral density [TMD] was significant in the middle [Ei = –7,399 + 9,520 * TMD, adjusted R2 = 0.34, P = 0.002] and proximal VOIs [Ei = –8,725 + 10,331 * TMD, adjusted R2 = 0.30, P = 0.004] but not in the distal VOI (P = 0.2).

than deeper subchondral bone with superficial subchondral bone being less stiff and capable of dissipating more energy than the adjacent deeper bone layer (Malekipour et al., 2018; Hargrave-Thomas et al., 2015; Choi et al., 1990). The lower stiffness in the distal VOI may limit this layer's contribution to fatigue resistance compared with the much stiffer middle and/or proximal VOIs (Malekipour et al., 2018). The positive association between bone volume fraction and stiffness in the distal VOI is contrary to a recent study where no associations between morphological parameters and stiffness were found in this location (Malekipour et al., 2018). Methodological differences in bone segmentation and stiffness measurements or the larger sample size in our study resulting in higher power could explain this discrepancy. Initial stiffness increased with decreasing volume fraction of highly mineralised bone in the distal VOI. This is consistent with the observation in our previous study of subchondral bone from the lateral metacarpal condyle where those horses with the highest bone volume fraction in the distal VOI also had the lowest volume fraction of highly mineralised bone in this location (Martig et al., 2018). This thin layer of bone immediately adjacent to the articular cartilage consistently has a lower tissue mineral density, possibly due to high bone turnover, which allows it to fulfil its function as an energy absorbing buffer between the overlying cartilage and the underlying stiffer bone (Boyde and Firth, 2005; Malekipour et al., 2018; Martig et al., 2018; Hargrave-Thomas

whether the association of such small changes in bone volume fraction with fatigue life persist with larger changes and when evaluating the entire bone. Higher bone volume fraction in the middle and proximal VOIs reflects a thicker subchondral bone plate, as is seen in horses in response to race training (Riggs and Boyde, 1999; Boyde and Firth, 2005). Our results suggest subchondral bone modelling in response to race training increases fatigue life and hence is a protective adaptation. Further investigations are required to reconcile this finding with the common collocation of fatigue damage with a thickened subchondral bone plate (Rubio-Martinez et al., 2008a; Muir et al., 2006; Whitton et al., 2018). The unexpected absence of an association between SCBI and fatigue life may be due to low power. We had a power of 26% to demonstrate an association between fatigue life and SCBI (f2 = 0.09, n = 22, 1 predictor), with a sample size of 93 specimens required to achieve 80% power with the same effect size. The low effect size could be due to the low number of specimens with severe SCBI. 4.2. Factors affecting initial stiffness Interestingly, bone volume fraction in the distal VOI was associated with initial stiffness but not fatigue life. Subchondral bone immediately adjacent to the articular cartilage has different mechanical properties 7

Journal of the Mechanical Behavior of Biomedical Materials 101 (2020) 103439

S. Martig, et al.

Fig. 4. Comparison of two different test ends for compressive fatigue testing of equine subchondral bone. Nf(Emin): number of cycles to failure defined as the cycle with the lowest recorded Young's modulus during cyclic compressive loading to 1.6 mm actuator displacement, corresponding to gross fracture. Nf (Emax-10%): number of cycles to 10% reduction of maximal recorded Young's modulus.

Interestingly, volume fraction of highly mineralised bone in the proximal but not the middle VOI was associated with increased initial stiffness. We previously found bone volume fraction was lower in the proximal than the middle VOI but volume fraction of highly mineralised bone was similar in the two VOIs (Martig et al., 2018). Thus a larger amount of bone is highly mineralised in the proximal compared to the middle VOI, which could lead to confounding with tissue mineral density. The absence of association between SCBI and initial stiffness is consistent with other studies, though it is somewhat counterintuitive (Rubio-Martinez et al., 2008b; Martig et al., 2013). Furthermore, we previously found that specimens with a lower volume fraction of highly mineralised bone in the distal VOI were less likely to have SCBI (Martig et al., 2018). In the current study, initial stiffness was higher in specimens with a lower volume fraction of highly mineralised bone in the distal VOI, hence an association between SCBI and initial stiffness could have been expected. The low number of horses with SCBI in general, and those with moderate and severe injury specifically, could explain the absence of an association between SCBI and initial stiffness. Post hoc power analysis revealed a power of 42% to detect an association between initial stiffness and SCBI (f2= 0.15, n = 23, 1 predictor). A sample size of 55 specimens would be required to reach a power of 80% to demonstrate an association of SCBI and stiffness assuming the same effect size. Moreover, gross assessment of SCBI does not account for the presence and extent of microscopic fatigue damage, nor that the damage is collocalised with the site of sample collection. Alternatively, given the relatively low effect size, other factors may counteract the presumed negative effect of SCBI on initial stiffness of these small bone specimens.

Fig. 3. Examples of changes of Young's modulus during compressive fatigue testing of equine subchondral bone (A) and stress-strain curves of a representative specimen (B). A: Empty symbols are for Young's modulus of cycles 1, 20, then every 20th cycle up to 100 cycles, followed by every 100th cycle to the end. For each of the three specimens the following are marked: cycle of maximal Young's modulus (full circle), cycle of 10% reduction of maximal Young's modulus (full triangle), cycle of 30% reduction of maximal Young's modulus (full square) and cycle of failure (full diamond). Note that Young's modulus at the cycle of failure in the specimen with less than 5500 cycles to failure (hollow squares) is > 70% of maximal Young's modulus. The specimen was grossly fractured at the cycle of failure. The full square for this specimen indicating 30% of maximal Young's modulus reduction is plotted at the same number of cycles of loading as the cycle of failure for illustration only. B: selected stress-strain curves of seven compressive loading cycles of the specimen represented in open triangles in (A). a: cycle 1, a short ramp envelope was used, thus the maximal load is less than 78 MPa; b: cycle 20; c: cycle 100; d: cycle of maximal Young's modulus (cycle 4900); e: cycle of 10% reduction of maximal Young's modulus (cycle 12,300); f: cycle of 30% reduction of maximal Young's modulus (cycle 14,800); g: cycle of lowest recorded Young's modulus (15,107).

et al., 2015). In the middle and proximal VOIs, higher initial stiffness was associated with higher tissue mineral density but not with bone volume fraction. The positive association between tissue mineral density and stiffness is consistent with findings in human cancellous bone (Follet et al., 2004). However, the absence of an association of bone volume fraction with stiffness is contrary to other reports where stiffness increased with increasing apparent density, a parameter that is highly influenced by bone volume fraction (Riggs and Boyde, 1999; Keaveny et al., 1997; Kopperdahl and Keaveny, 1998; Leahy et al., 2010). Methodological differences in the use of end constraint and the measurement of strain to calculate stiffness could explain this discrepancy.

4.3. Relationship of stiffness and fatigue life We found no association between initial stiffness and fatigue life. This is consistent with individual morphological parameters being associated with either initial stiffness or fatigue life. The strong association of maximal stiffness with fatigue life in the absence of an association of initial stiffness with fatigue life may be due to differences in strain rate between the two loading sequences as bone stiffness increases with strain rate (Dendorfer et al., 2008; Carter and Hayes, 1976). The strain rate for initial stiffness calculations was chosen to facilitate comparison of results with the literature and was lower than 8

Journal of the Mechanical Behavior of Biomedical Materials 101 (2020) 103439

S. Martig, et al.

Thomas et al., 2015). However, both chemical maceration and enzymatic cartilage removal require prolonged incubation in liquids at room or body temperature and hence are unsuitable prior to mechanical testing (Brown et al., 1993). The factors associated with subchondral bone stiffness and fatigue life possibly interact with each other. However, there was insufficient power to conduct multivariable analysis and assess interaction effects.

the strain rate during fatigue testing designed to mimic a galloping horse (Rubio-Martinez et al., 2008b; Witte et al., 2006; Seder and Vickery, 2003; Leahy et al., 2010). Post hoc power analysis revealed that we had a power of 52% (f2 = 0.20, n = 22, 1 predictor) to determine an association between initial stiffness and fatigue life and 41 specimens would be required to reach a power of 80%. 4.4. Definition of failure

5. Conclusion We found that the cycle of 10% reduction of maximal Young's modulus correlates well with our definition of the cycle of fatigue failure, which depends on actuator displacement at gross specimen fracture (Martig et al., 2013). Determination of maximal stiffness was equivocal in only one specimen that had an initial lower stiffness peak followed by a rapid loss of stiffness before maximum stiffness was reached after a steady increase. This specimen may have failed after the first peak despite losing less than 10% stiffness. Specimens with shorter fatigue lives in this study tended to fracture grossly with lower reductions of maximum stiffness (i.e. often before reaching a 30% reduction in maximum stiffness). Thus, it is possible that specimens with a shorter fatigue life than we observed may fracture grossly before reaching even a 10% maximum stiffness reduction. Once maximum stiffness was determined, the number of cycles to 10% reduction of maximal stiffness could unequivocally be found in all specimens. Conversely, determination of the cycle of failure based on the cycle of lowest recorded Young's modulus required further interpretation of the stiffness data in three of 22 specimens. Hence the number of cycles to 10% reduction of maximal stiffness is a valid alternative test end point independent of actuator displacement at gross specimen fracture to be considered for future studies. The pattern of change in stiffness during compressive fatigue testing was similar to that previously reported for equine subchondral bone (Martig et al., 2013). Cancellous bone also exhibits an increase in stiffness during cyclic compressive loading, though stiffness measurements reached a steady state after 10 cycles of loading to 50% of ultimate strength (Linde and Hvid, 1987). We did not determine monotonic mechanical properties of our specimens but the yield stress of similar specimens has been reported as 92 MPa (Rubio-Martinez et al., 2008b). Thus, the load of 78 MPa used in our study is higher than 50% of ultimate strength, which could contribute to the prolonged stiffness increase. Assessment of residual strain could be helpful to further investigate this behaviour. Residual strain could also be investigated as an alternative measure of damage accumulation and possible parameter to define the test end. Further research is required to investigate the underlying mechanism of increasing stiffness during compression-compression fatigue testing of subchondral bone and whether it is associated with reversible or irreversible structural changes.

We found that longer compressive fatigue life was associated with increased subchondral bone plate thickness in the palmar aspect of the metacarpal condyle, an area undergoing cyclic high loading in the equine athlete. This suggests adaptive modelling in response to race training has a protective effect on subchondral bone, which is at odds with theories that implicate densification of subchondral bone in the development of subchondral bone injury. Subchondral bone stiffness increased during cyclic compressive loading and it was the magnitude of the maximal stiffness reached that was associated with fatigue life rather than the initial subchondral bone stiffness. A longer fatigue life implies increased resistance to fatigue damage accumulation, hence further research is required to investigate the cause of the common collocation of fatigue damage and areas of increased bone volume fraction (Norrdin and Stover, 2006; Rubio-Martinez et al., 2008a; Muir et al., 2006; Whitton et al., 2018). Declaration of interest Sandra Martig was supported by an Australian Government Research Training Program Scholarship. The micro-CT examinations were supported by a grant from the ANZ Kathleen Agnes Back Estate. Peta Hitchens and Chris Whitton were supported by funding from Racing Victoria Limited, the Victorian Racing Industry Fund of the Victorian State Government, and The University of Melbourne. None of the funding bodies had any influence in study design; collection, analysis and interpretation of data; writing of the report; and the decision to submit the article for publication. The authors declare no other interests. Author contributions S.M., P.V.S.L and R.C.W. designed the study, S.M. acquired the data and performed the initial data analyses, S.M. and P.L.H. performed the final data analyses, S.M. drafted the manuscript and R.C.W. revised the manuscript critically for important intellectual content. All authors commented on draft versions of the manuscript and approved the final version of the manuscript.

4.5. Limitations Appendix A. Supplementary data Limitations of this study include the cross-sectional design, convenience sampling, reliance on trainer information and small sample size. Some trainers elected not to provide training information or could not be contacted, and those horses were not included in the study. This could have introduced selection bias but because this study focused on bone mechanical properties rather than attributes of the Victorian racehorse population we believe this bias is acceptable for the purpose of this study. The reliance on oral information provided retrospectively by the trainers is another weakness as the accuracy of such information cannot be controlled. Cross-sectional studies do not provide the longitudinal information required to determine the true effect of training on subchondral bone mechanical properties, but currently available longitudinal in vivo bone fatigue models are based on rodent models and are not transferable to horses. Further, removal of the non-calcified articular cartilage with a scalpel blade may have caused damage to the calcified cartilage, which could affect mechanical properties (Hargrave-

Supplementary data to this article can be found online at https:// doi.org/10.1016/j.jmbbm.2019.103439. References Barr, E.D., Pinchbeck, G.L., Clegg, P.D., Boyde, A., Riggs, C.M., 2009. Post mortem evaluation of palmar osteochondral disease (traumatic osteochondrosis) of the metacarpo/metatarsophalangeal joint in Thoroughbred racehorses. Equine Vet. J. 41, 366–371. Bowman, S.M., Guo, X.E., Cheng, D.W., Keaveny, T.M., Gibson, L.J., Hayes, W.C., et al., 1998. Creep contributes to the fatigue behavior of bovine trabecular bone. J. Biomech. Eng. Trans. ASME 120, 647–654. Boyde, A., Firth, E.C., 2005. Musculoskeletal responses of 2-year-old Thoroughbred horses to early training. 8. Quantitative back-scattered electron scanning electron microscopy and confocal fluorescence microscopy of the epiphysis of the third metacarpal bone. N. Z. Vet. J. 53, 123–132. Brown, R.A., Blunn, G.W., Salisbury, J.R., Byers, P.D., 1993. Two patterns of calcification

9

Journal of the Mechanical Behavior of Biomedical Materials 101 (2020) 103439

S. Martig, et al.

morphology in the metacarpus of racehorses in training changes with distance from the articular surface but not with age. J. Anat. 232, 919–930. Mazurkiewicz, A., Topolinski, T., 2017. Relationship between the mineral content of human trabecular bone and selected parameters determined from fatigue test at stepwise-increasing amplitude. Acta Bioeng. Biomech. 19, 19–26. Moore, T.L.A., Gibson, L.J., 2003. Fatigue of bovine trabecular bone. J. Biomech. Eng. Trans. ASME 125, 761–768. Muir, P., McCarthy, J., Radtke, C.L., Markel, M.D., Santschi, E.M., Scollay, M.C., et al., 2006. Role of endochondral ossification of articular cartilage and functional adaptation of the subchondral plate in the development of fatigue microcracking of joints. Bone 38, 342–349. Muir, P., Peterson, A.L., Sample, S.J., Scollay, M.C., Markel, M.D., Kalscheur, V.L., 2008. Exercise-induced metacarpophalangeal joint adaptation in the Thoroughbred racehorse. J. Anat. 213, 706–717. Norrdin, R.W., Stover, S.M., 2006. Subchondral bone failure in overload arthrosis: a scanning electron microscopic study in horses. J. Musculoskelet. Neuronal Interact. 6, 251–257. Otsu, N., 1979. Threshold selection method from gray-level histograms. IEEE Trans. Syst. Man Cybern. 9, 62–66. Pinchbeck, G.L., Clegg, P.D., Boyde, A., Riggs, C.M., 2013. Pathological and clinical features associated with palmar/plantar osteochondral disease of the metacarpo/ metatarsophalangeal joint in Thoroughbred racehorses. Equine Vet. J. 45, 587–592. Radin, E.L., Parker, H.G., Pugh, J.W., Steinber, Rs, Paul, I.L., Rose, R.M., 1973. Response of joints to impact loading - III. Relationship between trabecular microfractures and cartilage degeneration. J. Biomech. 6, 51–57. Rapillard, L., Charlebois, M., Zysset, P.K., 2006. Compressive fatigue behavior of human vertebral trabecular bone. J. Biomech. 39, 2133–2139. Riggs, C.M., Boyde, A., 1999. Effect of exercise on bone density in distal regions of the equine third metacarpal bone in 2-year-old thoroughbreds. Equine Vet. J. Suppl. S30, 555–560. Rubio-Martinez, L.M., Cruz, A.M., Gordon, K., Hurtig, M.B., 2008a. Structural characterization of subchondral bone in the distal aspect of third metacarpal bones from Thoroughbred racehorses via micro-computed tomography. Am. J. Vet. Res. 69, 1413–1422. Rubio-Martinez, L.M., Cruz, A.M., Gordon, K., Hurtig, M.B., 2008b. Mechanical properties of subchondral bone in the distal aspect of third metacarpal bones from Thoroughbred racehorses. Am. J. Vet. Res. 69, 1423–1433. Seder, J.A., Vickery, C.E.I., 2003. Temporal and kinematic gait variables of Thoroughbred racehorses at or near racing speeds. J. Equine Vet. Sci. 23, S82–S112. Topolinski, T., Cichanski, A., Mazurkiewicz, A., Nowicki, K., 2011. Study of the behavior of the trabecular bone under cyclic compression with stepwise increasing amplitude. J. Mech. Behav. Biomed. Mater. 4, 1755–1763. Turley, S.M., Thambyah, A., Riggs, C.M., Firth, E.C., Broom, N.D., 2014. Microstructural changes in cartilage and bone related to repetitive overloading in an equine athlete model. J. Anat. 224, 647–658. White, J., Yeats, A., Skipworth, G., 1977. Tables for Statisticians. Tables for Statisticians. University of Queensland Press, St.Lucia, Queensland, pp. 36–37. Whitton, R.C., Ayodele, B.A., Hitchens, P.L., Mackie, E.J., 2018. Subchondral bone microdamage accumulation in distal metacarpus of Thoroughbred racehorses. Equine Vet. J. 50, 766–773. Witte, T.H., Hirst, C.V., Wilson, A.M., 2006. Effect of speed on stride parameters in racehorses at gallop in field conditions. J. Exp. Biol. 209, 4389–4397.

in primary (physeal) and secondary (epiphyseal) growth cartilage. Clin. Orthop. Relat. Res. 318–324. Carter, D.R., Hayes, W.C., 1976. Bone compressive strength: the influence of density and strain rate. Science 194, 1174–1176. Choi, K., Kuhn, J.L., Ciarelli, M.J., Goldstein, S.A., 1990. The elastic moduli of human subchondral, trabecular, and cortical bone tissue and the size-dependency of cortical bone modulus. J. Biomech. 23, 1103–1113. Dendorfer, S., Maier, H.J., Hammer, J., 2006. Deformation behaviour of bovine cancellous bone. Technol. Health Care 14, 549–556. Dendorfer, S., Maier, H.J., Taylor, D., Hammer, J., 2008. Anisotropy of the fatigue behaviour of cancellous bone. J. Biomech. 41, 636–641. Fatihhi, S.J., Harun, M.N., Abdul Kadir, M.R., Abdullah, J., Kamarul, T., Ochsner, A., et al., 2015. Uniaxial and multiaxial fatigue life prediction of the trabecular bone based on physiological loading: a comparative study. Ann. Biomed. Eng. 43, 2487–2502. Faul, F., Erdfelder, E., Buchner, A., Lang, A.G., 2009. Statistical power analyses using G*Power 3.1: tests for correlation and regression analyses. Behav. Res. Methods 41, 1149–1160. Follet, H., Boivin, G., Rumelhart, C., Meunier, P.J., 2004. The degree of mineralization is a determinant of bone strength: a study on human calcanei. Bone 34, 783–789. Haddock, S.M., Yeh, O.C., Mummaneni, P.V., Rosenberg, W.S., Keaveny, T.M., 2004. Similarity in the fatigue behavior of trabecular bone across site and species. J. Biomech. 37, 181–187. Hargrave-Thomas, E., van Sloun, F., Dickinson, M., Broom, N., Thambyah, A., 2015. Multi-scalar mechanical testing of the calcified cartilage and subchondral bone comparing healthy vs early degenerative states. Osteoarthr. Cartil. 23, 1755–1762. Harrison, S.M., Whitton, R.C., Kawcak, C.E., Stover, S.M., Pandy, M.G., 2010. Relationship between muscle forces, joint loading and utilization of elastic strain energy in equine locomotion. J. Exp. Biol. 213, 3998–4009. Keaveny, T.M., Pinilla, T.P., Crawford, R.P., Kopperdahl, D.L., Lou, A., 1997. Systematic and random errors in compression testing of trabecular bone. J. Orthop. Res. 15, 101–110. Kopperdahl, D.L., Keaveny, T.M., 1998. Yield strain behavior of trabecular bone. J. Biomech. 31, 601–608. Lajeunesse, D., Massicotte, F., Pelletier, J.P., Martel-Pelletier, J., 2003. Subchondral bone sclerosis in osteoarthritis: not just an innocent bystander. Mod. Rheumatol. 13, 7–14. Leahy, P.D., Smith, B.S., Easton, K.L., Kawcak, C.E., Eickhoff, J.C., Shetye, S.S., et al., 2010. Correlation of mechanical properties within the equine third metacarpal with trabecular bending and multi-density micro-computed tomography data. Bone 46, 1108–1113. Li, G., Yin, J., Gao, J., Cheng, T.S., Pavlos, N.J., Zhang, C., et al., 2013. Subchondral bone in osteoarthritis: insight into risk factors and microstructural changes. Arthritis Res. Ther. 15, 223. Linde, F., Hvid, I., 1987. Stiffness behavior of trabecular bone specimens. J. Biomech. 20, 83–89. Malekipour, F., Whitton, C.R., Lee, P.V., 2018. Stiffness and energy dissipation across the superficial and deeper third metacarpal subchondral bone in Thoroughbred racehorses under high-rate compression. J. Mech. Behav. Biomed. Mater. 85, 51–56. Martig, S., Lee, P.V., Anderson, G.A., Whitton, R.C., 2013. Compressive fatigue life of subchondral bone of the metacarpal condyle in thoroughbred racehorses. Bone 57, 392–398. Martig, S., Hitchens, P.L., Stevenson, M.A., Whitton, R.C., 2018. Subchondral bone

10