### Summary

- Top of page
- Summary
- Introduction
- Materials and methods
- Results
- Discussion
- Acknowledgements
- Conflicts of interest
- Manufacturer's address
- References

**Reasons for performing study:** The flexor tendons support the metacarpophalangeal (MCP) and distal interphalangeal (DIP) joints during stance phase and since tendon stiffness and strain changes with age, it is likely that kinematics are also age-dependent.

**Hypothesis:** Maximum MCP and DIP angles decrease in the young horse, plateau in the mature horse and increase towards senescence.

**Methods:** The distal limbs of 57 walking horses age 3–212 months were filmed and digitised with an automated tracking system. Maximum MCP and DIP angles during stance phase were used to calculate strain in the superficial and deep digital flexor tendons. Horses were divided into 3 age groups; young (3–35 months), mature (36–99 months) and older horses (100–212 months). Pearson's correlation coefficients were calculated to determine the relationship between age and kinematics.

**Results:** Tendon strain decreased in young horses, stayed constant in mature horses and increased in older horses. Joint angles showed significant negative correlation in young horses, with coefficients of −0.88 (MCP) and −0.81 (DIP). In mature horses, correlations were not significant (P = 0.2 for MCP; P = 0.5 for DIP). In older horses, angles showed significant positive correlation, with coefficients of 0.62 (MCP) and 0.48 (DIP).

**Conclusions:** Joint angles decreased in the young horse as tendon stiffness increases, remained constant in the mature horse where tendon stiffness is constant and increased in older horses as tendons weakens and stiffness decreases. Strain patterns were similar to those found *in vitro*.

**Potential relevance:** Changing tendon stiffness appeared to influence the development and degeneration of gait. This has implications for studying musculoskeletal development, especially for identification of normal and pathological development.

### Introduction

- Top of page
- Summary
- Introduction
- Materials and methods
- Results
- Discussion
- Acknowledgements
- Conflicts of interest
- Manufacturer's address
- References

Tendon stiffness increases until it is optimal for the storage and release of elastic strain energy and therefore maximum locomotion efficiency in an individual. It has been proposed that tendon synthesis permanently ceases at skeletal maturity, between 24 and 36 months, once stiffness is optimal (Smith *et al*. 2002; Dowling and Dart 2005), resulting in minimal adaptive ability of the tendon beyond the age of about 36 months. Although certain exercise regimes have been found to optimise the structure of the tendon (Cherdchutham *et al*. 1999, 2001), these will only have an effect whilst matrix synthesis is ongoing. At present, determination of when matrix production has ceased is only possible through analysis of growth hormone (GH) concentration. Although urine analysis of GH concentration is possible, blood analysis is preferable due to the higher GH concentration (Popii and Baumann 2004, Bailly-Chouriberry *et al*. 2008). Blood collection is specialised and invasive.

After matrix production ceases, the minimal synthesis, coupled with repeated loading cycles, leads to a gradual degeneration of the tendon, manifest for example by focal and diffuse chondroid metaplasia after age 6 years (Crevier-Denoix *et al*. 1998). This degeneration leads to a decrease in mechanical integrity, such as a decline in concentration of strengthening cross-links after about age 15 years (Parry *et al*. 1978a; Gillis *et al*. 1997; Patterson-Kane *et al*. 1997a), loss of crimp in the collagen fibrils after about age 11 years (Buckwalter *et al*. 1993; Patterson-Kane *et al*. 1997a,b; Crevier-Denoix *et al*. 1998), a decrease in fascicle cross-sectionalarea after age 2 years (Gillis *et al*. 1997; Crevier-Denoix *et al*. 1998) and a decrease in COMP and collagen concentrations (Smith *et al*. 1997, 2002). The overall effect of this degeneration is a decrease in stiffness (Gillis *et al*. 1995). Since the average age of tendon degeneration onset is varied, a mean of 100 months was chosen for convenient statistical analysis.

The gradual decrease in mechanical integrity can result in the tendon no longer being able to withstand the loads placed on it (Smith *et al*. 2002) and being at risk of injury. The point at which the tendon is at risk depends on both the initial quality of the tendon and on the quantity of exercise given, and is thus subject-specific. Degeneration does not usually present clinical signs until the point of rupture and although ultrasound can be used to analyse tendon structure, this has not been found to be able to predict tendon injury (Avella *et al*. 2009).

A simple, noninvasive indicator of tendon mechanical properties could be the kinematics of the horse. The 3 joints of the distal limb are primarily supported during stance by the superficial (SDFT) and deep (DDFT) digital flexor tendons (Denoix 1994; McGuigan and Wilson 2003), although other structures such as the suspensory and sesamoidean ligaments have some influence. Maximum joint angle could, therefore, be expected to reflect tendon stiffness. Joint angles also depend on the mass of the horse, since the distal limb is passive, thus investigating the influence of tendon stiffness alone is achieved through normalisation to mass.

The objective of the study was to determine the specific effect of age on the stance phase maximum joint angles of the metacarpophalangeal (MCP) and distal interphalangeal (DIP), once mass had been accounted for, thus testing whether the effect of tendon stiffness on joint angle is detectable. The hypotheses were that in horses aged <36 months, whose tendon stiffness is assumed to be increasing, maximum joint angle will be negatively correlated to age. In mature horses aged 36–99 months, whose tendon stiffness is assumed to remain constant, maximum joint angle will have minimal correlation to age and in horses older than 100 months, whose tendon stiffness is assumed to be decreasing, maximum joint angle will be positively correlated to age. Strain in the SDFT and DDFT will be calculated from the joint angles in order to facilitate comparison of calculated strain to previously published *in vitro* data (Gillis *et al*. 1995).

### Materials and methods

- Top of page
- Summary
- Introduction
- Materials and methods
- Results
- Discussion
- Acknowledgements
- Conflicts of interest
- Manufacturer's address
- References

The distal forelimbs of 57 horses aged 3–212 months were filmed marker-free in walk at a self-selected speed. The population was of Sport Horse type of approximately similar build and included mares, stallions and castrated males. The population was of amateur standard and therefore subject to approximately similar exercise regimes. Horses were subject to regular clinical examination and showed no signs of lameness at data capture. Age was determined from the date of birth on the horses' passports. Two camera systems were compared to determine if 25 Hz is sufficient to detect age-related changes in locomotion: a Canon XL-H1 DV camera at 25 Hz and a Basler piA640-210gc at 100 Hz. To evaluate these systems, the maximum MCP and DIP angles recorded for the same 10 strides were compared by paired *t* test. The horses' line of progression was perpendicular to the camera at a distance of approximately 4 m and the field of view was approximately 3 m, sufficient to record at least one complete stride per trial. Since gait is dependant on velocity (Khumsap *et al*. 2002), velocities of the 3 groups were compared by ANOVA.

The mean sagittal plane joint angles for 3 complete stance phases were calculated for each horse. The videos were processed using novel proprietary automated marker-free tracking software written in MATLAB^{1} that tracked: the top of the metacarpus, the geometric centre of the MCP joint, the mid-point of the coronet band and the mid point of the bottom of the hoof (Fig 1). These coordinates were used to define segment vectors and trigonometry applied to calculate maximum values of MCP and DIP angles. The maximum values of the joint angles were chosen to replicate the conditions under which the SDFT and DDFT are at maximum length and, therefore, maximum strain. Since the DIP joint is contained within the hoof, an approximation of this joint was calculated using the geometric centre of the MCP, the mid-point of the coronet band and the mid-point of the hoof. The range of motion of the proximal interphalangeal joint (PIP) has been found to be only 10° (Chateau *et al*. 2004) and was, therefore, ignored in this study.

The automated marker-free tracking program consisted first of identification of the landmarks to be tracked. The 4 landmarks were manually digitised in the first frame and then tracked automatically using a combination of 3 constraints; firstly, all pixels whose intensity was similar to the initial pixels chosen were located; secondly, those pixels located further than a defined distance from the previous frame were eliminated; and finally, pixels were excluded that returned a segment length of ± 3% from the initial length to maintain the distal limb as a series of rigid segments.

If more than one pixel location remained, the average coordinates were returned. For each frame, the most distal point (bottom of the hoof) was found first because it was assumed to remain stationary during stance and thus show the smallest variability of location.

To validate the program, the precisely repeatable motion of a 4-bar linkage mechanism was tracked with both the new program and a widely accepted opto-electronic system and the angles reported by each system compared (Warlow *et al*. 2010). The new system reported an identical range of motion to the opto-electronic system and the standard deviation from a maximum angle of 123° was 0.6° across 10 trials. The maximum and minimum angles reported by each system differed by a maximum of 5°, possibly due to a difference in marker placement between systems. Although some limitations were demonstrated in tracking axial translation, this type of movement does not occur during stance phase and was, therefore, not considered a limitation for the current application.

An advantage of the new software was in eliminating some drawbacks of a marker-based system, such as skin displacement artefact and marker wobble, which can lead to inaccurate results (Mündermann *et al*. 2006). Although skin displacement artefact is low in the distal limb (van Weeren *et al*. 1988), discrepancies in distal limb joint angles have been found; for example a marker-based study suggested a DIP range of motion of 46° in walk (Clayton *et al*. 2007), whereas an invasive bone-pin study reported a range of motion of 59° (Chateau *et al*. 2004). Kinematic results from the present study were compared to results from this latter study for validation of the technique.

Since the mass of the horses could not be measured as no weighbridge was available, the abdominal depth in lateral view was digitised and used to indicate mass, which has been found to be highly correlated with abdominal size (Rodriguez *et al*. 2007). Two-dimensional abdominal depth was measured after exhalation from behind the withers to behind the elbow. To compensate for abdominal size being dependent on the distance from the camera, an object of known size was also digitised and used for calibration. To determine the influence of age on the relationship between abdominal depth and actual mass, the gradient of this relationship was compared between young and old horses. The gradient for young horses was calculated using data from a previously published study on a similar breed to the current study (Thompson 1995), aged 14 days–1.5 years. The older horses were aged 10–19 years.

The accuracy of this method was established by calculating the Pearson's correlation coefficient between digitised chest depth and actual mass of 13 horses. Maximum joint angle could then be normalised to mass to allow investigation of the specific effect of age.

Strain was calculated as a measure of stiffness in the SDFT and DDFT for comparison to previously published *in vitro* data (Gillis *et al*. 1995). Strain was calculated using a scaled link-segment model in MATLAB (Lawson *et al*. 2007), which took as input MCP and DIP angles and produced strains and an animation. The size and shape of the underlying skeleton were extracted from computed tomography scans and rendered as 3D objects in MATLAB. The paths of the SDFT and DDFT were overlaid on this skeleton, wrapping at both the proximal and distal sesamoid bones and at the proximal palmar extremity of the middle phalanx. MCP and DIP angles were input into the model, which then calculated tendon length and hence strain.

Both strain data and normalised joint angle data were calculated for 3 age groups: young horses aged 3–35 months (*Group 1*, n = 16); mature horses aged 36–99 months (*Group 2*, n = 20); and older horses aged 100–212 months (*Group 3*, n = 21). The cut-off of 36 months between *Groups 1* and *2* was chosen as the assumed age of tendon development completion, and the cut-off age of 100 months between *Groups 2* and *3* as the approximate mean of degeneration onset for statistical purposes. The influence of age on kinematics was analysed by 3 methods. Firstly, the Pearson's correlation coefficient was calculated for each joint angle and for strains in both the SDFT and DDFT. Secondly, assuming that the trend of joint angle to age can be modelled as linear in *Group 1* and *3* horses, and as a plateau in *Group 2*, the intersection of the trend of *Group 1* horses with the mean of *Group 2* was used to calculate the average age at which developmental changes appear to cease, and between *Group 2* and *3* horses to indicate the onset of degenerative change. Thirdly, the gradient of the slope of joint angle to age in *Group 1* and *3* horses was calculated to indicate rate of change.

### Results

- Top of page
- Summary
- Introduction
- Materials and methods
- Results
- Discussion
- Acknowledgements
- Conflicts of interest
- Manufacturer's address
- References

The tracking system provided a fast method of tracking anatomical landmarks throughout stance, processing each trial in <1 min. The automated nature of the program ensured its repeatability. The Pearson's correlation coefficient between calculated and actual mass was 0.81 (P = 0.001) and the r^{2} value was 0.65. The gradient of the relationship between girth depth and mass for young horses fell within one standard deviation of the gradient found in older horses, thus the influence of age on the chest depth-weight relationship was considered negligible.

There was no significant difference between the camera systems for either maximum MCP (P = 0.8) or maximum DIP (P = 0.5) angle. The range of speeds was 1.25–1.85 m/s and the mean value was 1.64 ± 0.03 m/s. There was no significant difference in walking speed between the 3 groups (P = 0.4). The range of motion was from 192–254° in the MCP, and 182–250° in the DIP.

As expected, strain showed significant negative correlation with age in *Group 1* horses (Figs 2, 3) with correlation coefficients of −0.81, P = 0.000 (SDFT) and −0.77, P = 0.001 (DDFT). In *Group 2* horses, coefficients were not significant for either SDFT (P = 0.6) or DDFT (P = 0.2) and the minimal gradient (0.3 and 0.7%,respectively) (Figs 2, 3) indicated no change with age. *Group 3* horses showed significant positive correlation of strain with age (Figs 2, 3), with correlation coefficients of 0.53, P = 0.01 (SDFT) and 0.54, P = 0.01 (DDFT).

Maximum MCP and DIP stance phase joint angle showed an initial significant decrease with age, a plateau and finally a significant increase towards old age (Figs 2, 3). Correlation coefficients for *Group 1* horses were −0.88, P = 0.000 (MCP) and −0.81, P = 0.000 (DIP), equivalent to a 1°/cm decrease over the group's age range of 2 years. In *Group 2* horses, coefficients were not significant for MCP (P = 0.2) or DIP (P = 0.5). The minimal gradient (0.5 and 0.2%, respectively) indicated no relationship with age (Figs 2, 3). *Group 3* horses showed significant positive correlation with age, with coefficients of 0.62, P = 0.003 (MCP) and 0.48, P = 0.03 (DIP), representing a rate of change of 0.5 and 0.2°/cm, respectively, over the group's age range of 9.3 years.

The intersection of the linear trend of joint angle in *Group 1* horses with the mean value of *Group 2* horses occurred at 29 (MCP) and 27 (DIP) months. The intersection between *Group 2* and *3* horses was found to occur at 136 (MCP) and 127 (DIP) months (Figs 2, 3).

### Discussion

- Top of page
- Summary
- Introduction
- Materials and methods
- Results
- Discussion
- Acknowledgements
- Conflicts of interest
- Manufacturer's address
- References

The aim of this study was to investigate the effect of age on the kinematics of equine distal limb joints, once mass had been accounted for. Fifty-seven horses were filmed in walk and the kinematics of the distal limb joints tracked with a novel automated marker-free tracking system. Joint angles were also used to calculate strain in the commonly-injured SDFT and DDFT.

Joint angles were found to be very similar to those found by invasive bone-pin methods (Chateau *et al*. 2004), indicating that the tracking system was accurate. Combined with its high throughput speed and repeatability, it is a potentially suitable tool for digitisation of video-based kinematics. The advantages of this system are that it can be used in strong sunlight that can interfere with infrared motion capture systems and it eliminates measurement inaccuracies due to marker wobble and skin displacement. The limitation of this system is that at present it is limited to 2D data.

The high correlation between actual and calculated mass indicated that the method employed was suitable for the breeds and age groups studied. The influence of age and breed on the relationship between calculated and actual mass was considered negligible for the breeds and age groups studied.

Joint angles showed a similar age-dependent change, showing an initial rapid decrease, followed by a plateau and a slower increase towards old age. The walking speed of the horses did not change significantly between the 3 age groups, indicating that the change in kinematics is independent of speed. In *Group 1*, the significant negative correlation is consistent with the hypothesis that increasing tendon stiffness could limit the maximum angle of the joints they wrap around. Contrary to the current study, maximum MCP angle has been found to increase from 4–26 months (Back *et al*. 1995). However, since these data were not normalised to mass, an increase in joint angle is expected as the increasing mass flexes the passive distal limb. Once normalised to the appropriate mass (Thompson 1995), maximum MCP decreased with age, consistent with the current study. In *Group 2* horses, once the change in tendon mechanical properties is assumed to have finished, the correlation with age was minimal in both the MCP and DIP. This implies that that age has little effect on the kinematics, and is consistent with the theory that there is little capacity for adaptive change in the mature tendon once synthesis has ceased (Buckwalter *et al*. 1993; Smith *et al*. 2002). In *Group 3* horses, the positive correlation of joint angles to age suggests that kinematics are again age-dependent in the ageing horse. This is in line with evidence that the ageing tendon starts to become more compliant as it degenerates (Gillis *et al*. 1995), which could allow an increase in maximum angle of the joints they support.

Whilst the distal limb joints are primarily supported by the flexor tendons (Denoix 1994; McGuigan and Wilson 2003), other structures such as the suspensory and sesamoidean ligaments have some influence. Tendon ageing is representative of the ageing of these other tissues; for example, the development and degeneration of the suspensory ligament is approximately parallel to that of the SDFT (Parry *et al*. 1978a). Despite the variability introduced by these other structures, the results suggest that the change in mechanical properties of the flexor tendons influences the development and degeneration of gait throughout the horse's life.

The age at intersection of the linear trend of *Group 1* with the plateau of *Group 2* was 29 (MCP) and 27 (DIP) months, although this will vary in the individual horse. This finding follows the theory that matrix synthesis production ceases between 24 and 36 months old once the tendon reaches optimal stiffness (Smith *et al*. 2002). The age at intersection of the plateau of *Group 2* with the linear trend of *Group 3* was 136 (MCP) and 127 months (DIP), consistent with the approximate mean of tendon degeneration onset at 100 months. These intersections provide evidence for the average age of division of the population.

In conclusion, kinematics changed with age; possibly influenced by the change in tendon stiffness throughout the horse's life. Joint angles showed a decrease with age in the young horse as tendon stiffness increased, remained constant in the mature horse whose tendon stiffness was constant, and increased in the older horse as stiffness decreased in the weakening tendon.

The first change in kinematics occurred at the age when matrix production ceases, so by studying an individual's kinematics from birth, an owner could identify this characteristic change, and therefore identify the limit of adaptive ability. Towards senescence, the degeneration of ageing tendon leads to an increase in its compliance and a second change in kinematics. This change is gradual and may be missed by traditional visual observation. The kinematics may allow detection of subtle kinematic changes before these become severe enough to alter the kinematics. This has benefits from both an equine welfare and economical point of view and also could have implications for the ageing human population.