Copyright is owned by the Author of the thesis. Permission is given for a copy to be downloaded by an individual for the purpose of research and private study only. The thesis may not be reproduced elsewhere without the permission of the Author. Predicting horse limb responses to surface variations with a 3D musculoskeletal model A thesis presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy In Biomechanics At Massey University, Manawatu, New Zealand Aliénor Bardin 2020 Predicting horse limb responses to surface variations i Abstract Thoroughbred racehorses are often affected by musculoskeletal injuries, leading to involuntary rest periods, early retirement or death. A number of studies have focused on identifying risk factors. A major focus of research has been track surface properties because it should be possible to modify these so that the risk of musculoskeletal injury is minimised. Among all the track surface properties studied to date, consistency of the surface is reported to be one of the main injury risk factors. The aim of this study was to develop a preliminary 3D musculoskeletal model of the whole equine forelimb based on data published in the literature and derived from anatomical measurements; and to determine the effects of the perturbations by the ground surface on the limb response with the musculoskeletal model developed and to assess whether the response occurs acutely in the perturbed stance phase or in the next stance phase. To answer these questions, gait data were collected from ridden Thoroughbreds passing through a perturbation area, where the surface hardness was changed by adding wood or foam under the baseline sand surface. The horses changed their joint flexion/extension patterns in response to changes in hardness. In response to the hard perturbation, the proximal limb spring was more compliant, evidenced by increased shoulder flexion. The elbow and carpal joints were more flexed in the intervening swing phase. In response to the soft perturbation, more coffin joint flexion was observed during both the perturbed and the following stance phase. The preliminary musculoskeletal model of the equine forelimb developed in this thesis allow the observation and study of the forelimb reaction to hardness perturbation through the joint excursions and tendon and ligament strains. Predicting horse limb responses to surface variations ii Acknowledgements First, I would like to thank Dr Bob Colborne, my main supervisor, for everything he did for me; for offering me the PhD position, for helping me with all the administrative tasks on my arrival in New Zealand, for helping me to find an accommodation, for teaching me the locomotion system of horses, for helping me to collect data, and for reading and correcting all I wrote. Thanks to Dr Chris Rogers, who helped me with writing, introduced me to other scientists at the ICEEP 2018 conference and allowed me to discover the NZ yearling parade. Thanks to my other two PhD committee members, Luca Panizzi, who gave good advice on my writing and to Dr Liqiong Tang, for her practical advice. I would like to express all my gratitude for the funding of this project and my PhD scholarship to the New Zealand Equine Trust. The Massey University Foundation and the Ron Kilgour Memorial Trust provided funding for my participation in the ICEEP 2018 conference in Australia. Many thanks to all the riders, horses and their owners who participated in the study and without whom it would not have been possible to run this study. I am grateful to all those who helped to collect data. Julie Huat and Astrid Meuly helped with the Thoroughbred limb dissection, Kylie Legg tested the ligament properties reported in the Appendix and Nila Taylor worked with us on an Equine Trust Summer Scholarship collecting and processing gait data. Finally, many thanks to my family, friends and acquaintances who supported me. Predicting horse limb responses to surface variations iii Table of contents Abstract .......................................................................................................................................... i Acknowledgements ....................................................................................................................... ii Table of contents .......................................................................................................................... iii List of figures ............................................................................................................................... viii List of tables ................................................................................................................................ xiii List of abbreviations .................................................................................................................... xvi Chapter 1 Introduction and literature review ............................................... 1 Introduction ........................................................................................................................... 1 Context .................................................................................................................................. 2 Racing industry ............................................................................................................... 2 Injuries in horse races..................................................................................................... 2 Modelling in biomechanics ............................................................................................. 5 Hoof-track interaction ........................................................................................................ 7 Risk factors ..................................................................................................................... 7 Surface properties .......................................................................................................... 7 Impactors ............................................................................................................... 8 Force plates and hoof-mounted devices ............................................................... 10 Comparison of surface properties ......................................................................... 12 Best properties? ................................................................................................... 16 Equine modelling ............................................................................................................. 19 The spring model .......................................................................................................... 19 Musculoskeletal model................................................................................................. 23 The different approaches ...................................................................................... 23 Reasons ................................................................................................................ 23 Equine musculoskeletal models ............................................................................ 25 Modelling challenges ............................................................................................ 27 Predicting horse limb responses to surface variations iv Conclusion ........................................................................................................................... 27 References ....................................................................................................................... 29 PART 1 CREATING THE MUSCULOSKELETAL MODEL .................................... 40 Chapter 2 Collecting Ligament and tendon properties ................................. 41 Introduction ......................................................................................................................... 41 Segment properties ............................................................................................................. 41 Muscle-tendon properties ................................................................................................ 42 Muscle-tendon properties required for modelling ........................................................ 42 Muscle-tendon data ..................................................................................................... 43 Ligament properties ......................................................................................................... 46 Conclusion and discussion .................................................................................................... 48 References ....................................................................................................................... 50 Chapter 3 Development of the musculoskeletal model ............................... 51 Introduction ......................................................................................................................... 51 Materials and methods ........................................................................................................ 51 Limb components ......................................................................................................... 51 The Anybody Modeling SystemTM ................................................................................. 53 Development of the model .............................................................................................. 54 Segments and initial positions ...................................................................................... 54 Joints, markers and kinematics ..................................................................................... 56 Scaling functions .......................................................................................................... 57 Changing the segment definitions ................................................................................ 61 Kinetic study................................................................................................................. 62 Final version ................................................................................................................. 64 Conclusion & Discussion ................................................................................................... 65 References ........................................................................................................................... 67 Chapter 4 Testing the preliminary model with gait data .............................. 69 Predicting horse limb responses to surface variations v Introduction ......................................................................................................................... 69 Method & Material .............................................................................................................. 69 Experimental setup ...................................................................................................... 69 Markers........................................................................................................................ 70 Results ............................................................................................................................. 71 Kinematic files .............................................................................................................. 72 Trot files ............................................................................................................... 72 Right lead canter files ........................................................................................... 72 Left lead canter files ............................................................................................. 73 Sagittal plane joint movements .................................................................................... 73 Coffin joint excursion ............................................................................................ 74 Fetlock joint excursion .......................................................................................... 75 Carpal joint excursion ........................................................................................... 78 Elbow joint excursion ............................................................................................ 80 Shoulder joint excursion ....................................................................................... 83 Tendon and ligament strains ........................................................................................ 84 Conclusion & Discussion ................................................................................................... 85 References ........................................................................................................................... 89 PART 2 STUDY OF THE HORSE LIMB RESPONSE TO GROUND HARDNESS PERTURBATION ......................................................................................... 90 Chapter 5 Gait data collection with perturbations ...................................... 91 Introduction ......................................................................................................................... 91 Material & Method .............................................................................................................. 91 Experimental set-up ..................................................................................................... 91 Perturbation pit ............................................................................................................ 92 Horses .......................................................................................................................... 92 Qualisys software ......................................................................................................... 94 Predicting horse limb responses to surface variations vi Results ............................................................................................................................. 96 Trot files ....................................................................................................................... 96 Canter files ................................................................................................................... 98 Discussion & Conclusion ................................................................................................... 99 References ......................................................................................................................... 102 Chapter 6 Limb response to perturbations................................................ 103 Introduction ....................................................................................................................... 103 Material & Methods........................................................................................................... 103 Results ........................................................................................................................... 104 Statistical results of the joint excursions ..................................................................... 104 Coffin joint: sagittal angle ................................................................................... 104 Fetlock joint: sagittal angle ................................................................................. 110 Carpal joint: sagittal angle................................................................................... 114 Elbow joint: sagittal angle ................................................................................... 118 Shoulder joint: sagittal angle............................................................................... 122 Statistical results of the ligament and tendon strains .................................................. 127 Ligament and tendon strains during stance at trot .............................................. 128 Ligament and tendon strains during stance at canter .......................................... 129 Summary of joint excursions and soft tissue strains ........................................................ 131 Conclusion ......................................................................................................................... 133 Response to perturbations at trot .............................................................................. 134 Response to perturbations at right lead canter ........................................................... 135 Discussion ...................................................................................................................... 136 References ..................................................................................................................... 139 Chapter 7 Conclusions ............................................................................. 140 Introduction ....................................................................................................................... 140 Effect of the perturbation .................................................................................................. 140 Predicting horse limb responses to surface variations vii Statistical model ......................................................................................................... 140 Responses to perturbations ........................................................................................ 141 Comparison to literature ............................................................................................ 144 Musculoskeletal model .................................................................................................. 145 Segments ................................................................................................................... 145 Ligament and muscle-tendon units ............................................................................. 145 Joints.......................................................................................................................... 146 Scaling ........................................................................................................................ 147 Model validation ........................................................................................................ 147 Conclusion ..................................................................................................................... 148 Hardness perturbation in racetrack ............................................................................ 148 Improving the musculoskeletal model ........................................................................ 149 References ......................................................................................................................... 151 Appendix: Ligament, muscle and tendon properties ................................. 153 Ligament properties ........................................................................................................... 153 Muscle-tendon data ........................................................................................................... 154 Muscle-tendon properties .......................................................................................... 154 Muscle-tendon origin and insertion sites .................................................................... 155 Lateral digital extensor (LDE) .............................................................................. 155 Common digital extensor (CDE) .......................................................................... 155 Extensor carpi obliquus (ECO) ............................................................................. 156 Extensor carpi radialis (ECR) ................................................................................ 157 Ulnaris lateralis (UL)............................................................................................ 158 Flexor carpi radialis (FCR) .................................................................................... 159 Flexor carpi ulnaris (FCU) .................................................................................... 160 Deep digital flexor (DDF) ..................................................................................... 161 Superficial digital flexor (SDF) ............................................................................. 161 Predicting horse limb responses to surface variations viii List of figures Figure 1 Distribution of the reported events with number of cases (Perkins et al., 2004b) (MSI = Musculoskeletal Injuries) ................................................................................................................... 3 Figure 2 CT derived image of the Scapula segment with its local reference frame on the estimated centre of mass. Lateral view on the left, cranial view on the right. ................................................... 55 Figure 3 Position of markers (green points) and antebrachiocarpal joint point (red point) on the antebrachium segment. Lateral (on the left) and cranial (on the right) views of the radius-ulna 3D image. All three points are superimposed on the lateral view .......................................................... 56 Figure 4 Dorsal (A), ventro-caudal (B) and medial (C) views of the elbow with the brachium and antebrachium scaled with the first scaling function. ........................................................................ 58 Figure 5 Lateral (A) and dorso-caudal (B) views of the distal phalanx scaled by the second scaling function........................................................................................................................................... 58 Figure 6 Dorsal (A), ventro-caudal (B) and medial (C) views of the elbow and lateral (D) and dorso-caudal (E) views of the distal phalanx with all segments scaled by the third scaling function.. 59 Figure 7 Problem of joint geometry due to scaling. Medial views of the elbow of the CT-scanned bones (on the left) and of the scaled segments (on the right) .......................................................... 60 Figure 8 Medial view of the elbow with the segments and joint geometries scaled by the final scaling function........................................................................................................................................... 61 Figure 9 Lateral (on the left) and cranial (on the right) of the distal phalanx with the contact points (red points) ..................................................................................................................................... 63 Figure 10 Diagram of the model ...................................................................................................... 64 Figure 11 Picture of the camera set-up used to collect gait data (on top) and diagram of the camera set-up (on bottom). The horse forelimb was aimed to land in the “forelimb landing zone” represented on the diagram, which match the zone just after the poles on the ground. .................. 70 Figure 12 Picture of a horse equipped with the whole set of markers used to collect the Initialization data ................................................................................................................................................. 71 Figure 13 Example of coffin angles. Joint angle is defined as the angle between the long axis of the metacarpus segment and the long axis of the pastern segment, as indicated. The positive sign corresponds to a flexed position and the negative sign to an extended position. ............................. 74 Figure 14 Right forelimb coffin joint excursions at trot (A), right lead canter (B) and left lead canter (C). The graphs present the average and standard deviation computed from all the gait data files (n=53 files for trot, n=23 for right lead canter and n=21 files for left lead canter). The data were time- normalized from -1 to 0 as stance phase and from 0 to 1 as swing phase. ........................................ 74 Predicting horse limb responses to surface variations ix Figure 15 Example of fetlock angles. Joint angle is defined as the angle between the long axis of the metacarpus segment and the long axis of the pastern segment, as indicated. The positive sign corresponds to a flexed position and the negative sign to an extended position. ............................. 75 Figure 16 Right forelimb fetlock joint excursions at trot (A), right lead canter (B) and left lead canter (C). The graphs present the average and standard deviation computed from all the gait data files (n=53 files for trot, n=23 for right lead canter and n=21 files for left lead canter). The data were time- normalized from -1 to 0 as stance phase and from 0 to 1 as swing phase. ........................................ 76 Figure 17 Example of carpal angles. Joint angle is defined as the angle between the long axis of the antebrachium segment and the long axis of the metacarpus segment, as indicated. The positive sign corresponds to a flexed position. ..................................................................................................... 78 Figure 18 Right forelimb carpal joint excursions at trot (A), right lead canter (B) and left lead canter (C). The graphs present the average and standard deviation computed from all the gait data files (n=53 files for trot, n=23 for right lead canter and n=21 files for left lead canter). The data were time- normalized from -1 to 0 as stance phase and from 0 to 1 as swing phase. ........................................ 78 Figure 19 Example of elbow joint angles. Joint angle is defined as the angle between the long axis of the brachium segment and the long axis of the antebrachium segment, as indicated. The positive sign corresponds to a flexed position. .............................................................................................. 80 Figure 20 Right forelimb elbow joint excursions at trot (A), right lead canter (B) and left lead canter (C). The graphs present the average and standard deviation computed from all the gait data files (n=53 files for trot, n=23 for right lead canter and n=21 files for left lead canter). The data were time- normalized from -1 to 0 as stance phase and from 0 to 1 as swing phase. ........................................ 81 Figure 21 Example of shoulder joint angles. Joint angle is defined as the angle between the long axis of the scapula segment and the long axis of the brachium segment, as indicated. The positive sign corresponds to a flexed position. ..................................................................................................... 83 Figure 22 Right forelimb shoulder joint excursions at trot (A), right lead canter (B) and left lead canter (C). The graphs present the average and standard deviation computed from all the gait data files (n=53 files for trot, n=23 for right lead canter and n=21 files for left lead canter). The data were time-normalized from -1 to 0 as stance phase and from 0 to 1 as swing phase. ............................... 83 Figure 23 Diagram illustrating the filling of the perturbation pit to create the perturbation. The hole is filled with 12cm of wood and covered with 6cm of sand for the baseline (on the left); with 5cm of sand in the bottom and 12cm of wood and covered with 1cm of sand for the hard perturbation (in the middle); and with 12cm of wood covered with 5cm of foam and 1cm of sand for the soft perturbation (on the right). ............................................................................................................. 92 Predicting horse limb responses to surface variations x Figure 24 Horse forelimb equipped with the segmental and joint markers to collect the Initialization file. Cranio-lateral view (on the left) and lateral view (in the middle) of the whole forelimb and cranio-lateral view of the distal forelimb (from carpus to hoof) on the right .................................... 93 Figure 25 Snapshot of the marker positions on the whole forelimb at the beginning of the stance phase for a trial at trot with Qualisys Track Manager software ........................................................ 94 Figure 26 Example of trajectory with gaps to fill (on the left) and zoom on one gap being filled (on the right) using the Qualisys interface ............................................................................................. 95 Figure 27 Coffin joint excursion at trot over the baseline (BL, on top), over the hard perturbation (HP, in the middle) and over the soft perturbation (SP, on bottom). The curves represent the averages (plain lines) and the standard error of mean (dotted lines) of the coffin joint excursion at trot computed for all the files (n=32 files for BL; n=27 files for HP and n=24 files for SP). Time was normalized from -1 to 0 as the perturbed stance phase, from 0 to 1 as the swing phase and from 1 to 2 as the second stance phase. ....................................................................................................... 105 Figure 28 Coffin joint excursion at right lead canter over the baseline (BL, on top), over the hard perturbation (HP, in the middle) and over the soft perturbation (SP, on bottom). The curves represent the averages (plain lines) and the standard error of mean (dotted lines) of the coffin joint excursion at canter computed for all the files (n=24 files for BL; n=28 files for HP and n=24 files for SP). Time was normalized from -1 to 0 as the perturbed stance phase, from 0 to 1 as the swing phase and from 1 to 2 as the second stance phase. ................................................................................. 108 Figure 29 Fetlock joint excursion at trot over the baseline (BL, on top), over the hard perturbation (HP, in the middle) and over the soft perturbation (SP, on bottom). The curves represent the averages (plain lines) and the standard error of mean (dotted lines) of the fetlock joint excursion at trot computed for all the files (n=32 files for BL; n=27 files for HP and n=24 files for SP). Time was normalized from -1 to 0 as the perturbed stance phase, from 0 to 1 as the swing phase and from 1 to 2 as the second stance phase. ....................................................................................................... 111 Figure 30 Fetlock joint excursion at right lead canter over the baseline (BL, on top), over the hard perturbation (HP, in the middle) and over the soft perturbation (SP, on bottom). The curves represent the averages (plain lines) and the standard error of mean (dotted lines) of the fetlock joint excursion at canter computed for all the files (n=24 files for BL; n=28 files for HP and n=24 files for SP). Time was normalized from -1 to 0 as the perturbed stance phase, from 0 to 1 as the swing phase and from 1 to 2 as the second stance phase. ................................................................................. 113 Figure 31 Carpal joint excursion at trot over the baseline (BL, on top), over the hard perturbation (HP, in the middle) and over the soft perturbation (SP, on bottom). The curves represent the averages (plain lines) and the standard error of mean (dotted lines) of the carpal joint excursion at Predicting horse limb responses to surface variations xi trot computed for all the files (n=32 files for BL; n=27 files for HP and n=24 files for SP). Time was normalized from -1 to 0 as the perturbed stance phase, from 0 to 1 as the swing phase and from 1 to 2 as the second stance phase. ....................................................................................................... 115 Figure 32 Carpal joint excursion at right lead canter over the baseline (BL, on top), over the hard perturbation (HP, in the middle) and over the soft perturbation (SP, on bottom). The curves represent the averages (plain lines) and the standard error of mean (dotted lines) of the carpal joint excursion at canter computed for all the files (n=24 files for BL; n=28 files for HP and n=24 files for SP). Time was normalized from -1 to 0 as the perturbed stance phase, from 0 to 1 as the swing phase and from 1 to 2 as the second stance phase. ................................................................................. 117 Figure 33 Elbow joint excursion at trot over the baseline (BL, on top), over the hard perturbation (HP, in the middle) and over the soft perturbation (SP, on bottom). The curves represent the averages (plain lines) and the standard error of mean (dotted lines) of the elbow joint excursion at trot computed for all the files (n=32 files for BL; n=27 files for HP and n=24 files for SP). Time was normalized from -1 to 0 as the perturbed stance phase, from 0 to 1 as the swing phase and from 1 to 2 as the second stance phase. ....................................................................................................... 119 Figure 34 Elbow joint excursion at right lead canter over the baseline (BL, on top), over the hard perturbation (HP, in the middle) and over the soft perturbation (SP, on bottom). The curves represent the averages (plain lines) and the standard error of mean (dotted lines) of the elbow joint excursion at canter computed for all the files (n=24 files for BL; n=28 files for HP and n=24 files for SP). Time was normalized from -1 to 0 as the perturbed stance phase, from 0 to 1 as the swing phase and from 1 to 2 as the second stance phase. ................................................................................. 121 Figure 35 Shoulder joint excursion at trot over the baseline (BL, on top), over the hard perturbation (HP, in the middle) and over the soft perturbation (SP, on bottom). The curves represent the averages (plain lines) and the standard error of mean (dotted lines) of the shoulder joint excursion at trot computed for all the files (n=32 files for BL; n=27 files for HP and n=24 files for SP). Time was normalized from -1 to 0 as the perturbed stance phase, from 0 to 1 as the swing phase and from 1 to 2 as the second stance phase. ....................................................................................................... 123 Figure 36 Shoulder joint excursion at right lead canter over the baseline (BL, on top), over the hard perturbation (HP, in the middle) and over the soft perturbation (SP, on bottom). The curves represent the averages (plain lines) and the standard error of mean (dotted lines) of the shoulder joint excursion at canter computed for all the files (n=24 files for BL; n=28 files for HP and n=24 files for SP). Time was normalized from -1 to 0 as the perturbed stance phase, from 0 to 1 as the swing phase and from 1 to 2 as the second stance phase. ....................................................................... 126 Figure 37 Lateral view of the radius/ulna. Radial origin sites of the LDE ......................................... 155 Predicting horse limb responses to surface variations xii Figure 38 Lateral view of the distal part of the humerus. Humeral origin site of the CDE................ 156 Figure 39 Lateral view of the proximal part of the radius/ulna. Radial origin of the CDE ................. 156 Figure 40 Lateral view of the radius/ulna. Origin site of the ECO .................................................... 156 Figure 41 Cranial view of the distal part of the radius. Origin site of the ECO ................................. 156 Figure 42 Lateral view of the proximal part of the metacarpus. Insertion site of the ECO ............... 157 Figure 43 Lateral view of the distal part of the humerus. Humeral origin of the ECR ...................... 157 Figure 44 Drawing of the UL and the sites of samples used to measure the UL properties (in orange) ...................................................................................................................................................... 158 Figure 45 Lateral view of the distal part of the humerus. Origin site of the UL ............................... 159 Figure 46 Lateral view of the carpus. Insertion sites of the UL ........................................................ 159 Figure 47 Caudal view of the carpus. Insertion sites of the UL ........................................................ 159 Figure 48 Medial view of the distal part of the humerus. Origin site of the FCR.............................. 160 Figure 49 Caudal view of the carpus. Insertion site of the FCR ....................................................... 160 Figure 50 Medial view of the distal part of the humerus. Origin site of the FCU ............................. 160 Figure 51 Caudal view of the carpus. Insertion site of the FCU ....................................................... 160 Figure 52 Medial view of the radius/ulna. Radial and ulnar insertion sites of the DDF .................... 161 Figure 53 Medial view of the distal pert of the humerus. Humeral origin site of the DDF ............... 161 Figure 54 Medial view of the distal part of radius/ulna. Radial origin of the SDF ............................ 162 Predicting horse limb responses to surface variations xiii List of tables Table 1 Muscle properties that are required or that can be added for each muscle model............... 43 Table 2 Muscle data used for the AnyBody Thoroughbred forelimb model, obtained from published literature (1-5) and from laboratory dissection data (6) ................................................................... 46 Table 3 Ligament data used for modelling which was obtained from published literature ................ 48 Table 4 Summary of the models published in the literature, the model used in this study and the key differences between these models .................................................................................................. 52 Table 5 Trot data files, details per horse: number of files, velocity, stance phase duration and swing phase duration (average ± standard deviation). ............................................................................... 72 Table 6 Right lead canter data files, details per horse: number of files, velocity, stance phase duration and swing phase duration (average ± standard deviation). ................................................ 73 Table 7 Left lead canter data files, details per horse: number of files, velocity, stance phase duration and swing phase duration (average ± standard deviation). .............................................................. 73 Table 8 Right forelimb coffin flexion and extension peaks at trot. Amplitude and time of occurrence reported in the literature and computed in the model (average ± standard deviation). The data were time-normalized from -1 to 0 as stance phase and from 0 to 1 as swing phase. ............................... 75 Table 9 Right forelimb fetlock flexion and extension peaks at trot. Amplitude and time of occurrence reported in the literature and computed in the model (average ± standard deviation). The data were time-normalized from -1 to 0 as stance phase and from 0 to 1 as swing phase. ............................... 76 Table 10 Right forelimb fetlock flexion and extension peaks at right lead canter. Amplitude and time of occurrence reported in the literature and computed in the model (average ± standard deviation). The data were time-normalized from -1 to 0 as stance phase and from 0 to 1 as swing phase. ........ 77 Table 11 Right forelimb fetlock flexion and extension peaks at left lead canter. Amplitude and time of occurrence reported in the literature and computed in the model (average ± standard deviation). The data were time-normalized from -1 to 0 as stance phase and from 0 to 1 as swing phase. ........ 77 Table 12 Right forelimb carpal flexion and extension peaks at trot. Amplitude and time of occurrence reported in the literature and computed in the model (average ± standard deviation). The data were time-normalized from -1 to 0 as stance phase and from 0 to 1 as swing phase. ............................... 79 Table 13 Right forelimb carpal flexion and extension peaks at right lead canter. Amplitude and time of occurrence reported in the literature and computed in the model (average ± standard deviation). The data were time-normalized from -1 to 0 as stance phase and from 0 to 1 as swing phase. ........ 79 Table 14 Right forelimb carpal flexion and extension peaks at left lead canter. Amplitude and time of occurrence reported in the literature and computed in the model (average ± standard deviation). The data were time-normalized from -1 to 0 as stance phase and from 0 to 1 as swing phase. ........ 80 Predicting horse limb responses to surface variations xiv Table 15 Right forelimb elbow flexion and extension peaks at trot. Amplitude and time of occurrence reported in the literature and computed in the model (average ± standard deviation). The data were time-normalized from -1 to 0 as stance phase and from 0 to 1 as swing phase. ............................... 81 Table 16 Right forelimb elbow flexion and extension peaks at right lead canter. Amplitude and time of occurrence reported in the literature and computed in the model (average ± standard deviation). The data were time-normalized from -1 to 0 as stance phase and from 0 to 1 as swing phase. ........ 82 Table 17 Right forelimb elbow flexion and extension peaks at left lead canter. Amplitude and time of occurrence reported in the literature and computed in the model (average ± standard deviation). The data were time-normalized from -1 to 0 as stance phase and from 0 to 1 as swing phase. ........ 82 Table 18 Trot data files, details per horse: number of files, velocity, 1st and 2nd stance phase durations and swing phase duration (average ± standard deviation) for each perturbation condition ........................................................................................................................................................ 97 Table 19 Right lead canter data files, details per horse: number of files, velocity, 1st and 2nd stance phase durations and swing phase duration (average ± standard deviation) for each perturbation condition ......................................................................................................................................... 99 Table 20 Coffin joint excursion at trot: peak values and their time of occurrence over the baseline, the hard and the soft perturbation conditions (average ± SEM). .................................................... 106 Table 21 Coffin joint excursion at right lead canter: peak values and their time of occurrence over the baseline, the hard and the soft perturbation conditions (average ± SEM). ..................................... 109 Table 22 Fetlock joint excursion at trot: peak values and their time of occurrence over the baseline, the hard and the soft perturbation conditions (average ± SEM). .................................................... 112 Table 23 Fetlock joint excursion at canter: peak values and their time of occurrence over the baseline, the hard and the soft perturbation conditions (average ± SEM). ..................................... 114 Table 24 Carpal joint excursion at trot: peak values and their time of occurrence over the baseline, the hard and the soft perturbation conditions (average ± SEM). .................................................... 116 Table 25 Carpal joint excursion at canter: peak values and their time of occurrence over the baseline, the hard and the soft perturbation conditions (average ± SEM). .................................................... 118 Table 26 Elbow joint excursion at trot: peak values and their time of occurrence over the baseline, the hard and the soft perturbation conditions (average ± SEM). .................................................... 120 Table 27 Elbow joint excursion at canter: peak values and their time of occurrence over the baseline, the hard and the soft perturbation conditions (average ± SEM). .................................................... 122 Table 28 Shoulder joint excursion at trot: peak values and their time of occurrence over the baseline, the hard and the soft perturbation conditions (average ± SEM). .................................................... 124 Predicting horse limb responses to surface variations xv Table 29 Shoulder joint excursion at canter: peak values and their time of occurrence over the baseline, the hard and the soft perturbation conditions (average ± SEM). ..................................... 127 Table 30 Maximal strains computed within the tendons and ligaments at trot during the perturbed stance phase and the following stance phase over the perturbation conditions (average ± SEM). .. 128 Table 31 Maximal strains computed within the tendons and ligaments at canter during the perturbed stance phase and the following stance phase over the perturbation conditions (average ± SEM). Horse and velocity differences were accounted for and the effect of velocity were removed in the averages presented here ............................................................................................................... 130 Table 32 Main effects of the hard perturbation on the joint excursions and tendon and ligament strains at trot ................................................................................................................................ 131 Table 33 Main effects of the hard perturbation on the joint excursions and tendon and ligament strains at canter ............................................................................................................................ 132 Table 34 Main effects of the soft perturbation on the joint excursions and tendon and ligament strains at trot ................................................................................................................................ 132 Table 35 Main effects of the soft perturbation on the joint excursions and tendon and ligament strains at canter ............................................................................................................................ 133 Table 36 Ligament properties ........................................................................................................ 153 Table 37 Muscle and tendon properties......................................................................................... 154 Predicting horse limb responses to surface variations xvi List of abbreviations AIC Akaike Information Criterion ALDDF Accessory Ligament of the Deep Digital Flexor ALSDF Accessory Ligament of the Superficial Digital Flexor ANOVA ANalysis Of Variance BL BaseLine CDE Common Digital Extensor CT Computed Tomography DDF Deep Digital Flexor ECO Extensor Carpi Obliquus ECR Extensor Carpi Radialis FCR Flexor Carpi Radialis FCU Flexor Carpi Ulnaris HP Hard Perturbation LDE Lateral Digital Extensor MSI MusculoSkeletal Injuries SDF Superficial Digital Flexor SDFT Tendon of the Superficial Digital Flexor SEM Standard Error of the Mean SP Soft Perturbation St-swing Stance-swing phase transition UL Ulnaris Lateralis Predicting horse limb responses to surface variations 1 Chapter 1 Introduction and literature review Introduction The Thoroughbred industry has an important place in the New Zealand economy accounting for approximately 1% of New Zealand’s Gross Domestic Product (Bolwell et al., 2017a). However, many of the Thoroughbreds that enter race training will be affected by musculoskeletal injuries, which would lead to rest periods in the best cases, but could also lead to retirement or death. The most frequently reported location for the musculoskeletal injuries are the distal forelimb (Perkins et al., 2004b). A number of studies have focused on identifying risk factors. One of the major areas of research is track surface properties, because it should be possible to modify these so that the risk of musculoskeletal injury would be reduced (Oikawa et al., 2000; Setterbo et al., 2013). Among all the track surface properties examined, consistency was reported to be one of the main injury risk factors. However, to date, only epidemiological studies have focused on this property. The aim of this thesis is therefore to test how unexpected variations in surface hardness, leading to inconsistency in track surface properties, affect the motion and loading of the limb, which could lead to an increased risk of musculoskeletal injury. For that, a model is required. Different equine limb models have been developed, and the most developed is the spring-mass model (McGuigan and Wilson, 2003). However, to study the effects of inconsistency of a track on the soft tissues of the limb requires the use of a model that includes those soft tissues and therefore a musculoskeletal model is needed. Other equine musculoskeletal models have been developed, but they generally have only included the distal limb (Brown et al., 2003b; Symons et al., 2016). The musculoskeletal model requires input parameters derived from gait and force plate data against which the model can be tested and developed. To obtain these data the ground hardness was altered in one discrete area of an indoor riding arena by adding wood or foam under the sand surface, to perturb the horse’s gait during one stance phase. Gait data were collected from horses passing through the perturbed area and for the subsequent swing and stance phases. The main objectives of this thesis were twofold: Chapter 1 Literature review 2 - To develop a preliminary 3D musculoskeletal model of the whole equine forelimb based on data published in the literature and derived from anatomical measurements (Part 1) - To determine the effects of the perturbations by the ground surface on the limb response with the musculoskeletal model developed and to assess whether the response occurs acutely in the perturbed stance phase or in the next stance phase (Part 2) Context Racing industry The history of the Thoroughbred in New Zealand began in 1840 when the first Thoroughbred stallion was imported. That same year, the first race meeting was organised in Wellington. Since then, the Thoroughbred industry has expanded and is now the most economically important part of the New Zealand equine industry, with approximately 40% of the annual Thoroughbred foal crop exported (Rogers et al., 2017). Thoroughbred and Standardbred racing and the sport-horse industries generate around 2% of New Zealand’s GDP (Gross Domestic Product) (Bolwell et al., 2017a). Regarding the international Thoroughbred industry, New Zealand has the 6th largest breeding industry (Gee et al., 2017) and is the 11th largest racing jurisdiction based on the number of horses starting in races (Bolwell et al., 2017a). Every year, around 5,500 and 300 horses start in approximately 2,900 flat races and 120 jump races respectively (Bolwell et al., 2016, 2017a; Rogers et al., 2017). Within New Zealand, since the global financial crisis (2007-2008), there has been a consistent reduction of the number of Thoroughbred foals born but the number of horses exported and the proportion of horses lost from the racing industry has remained relatively unchanged (Rogers et al., 2014). Horses may be lost from the racing industry because of voluntary or involuntary reasons. About one-third of Thoroughbreds that enter race training are retired prematurely, mainly because they lack talent, and another third involuntarily, in 78% of these cases because of musculoskeletal injuries (Perkins et al., 2004b). Thus, musculoskeletal injury represents the most significant, potentially manageable, reason for horses being lost from training and the industry. Injuries in horse races The importance of horses involuntarily retired, or that had reduced training or racing opportunities due to injury, is highlighted by the number of epidemiological studies that have been published. Some of these studies have focused on events occurring during racing. However, it has been reported that these race day events account for only a small proportion of total injuries in Thoroughbreds and that Predicting horse limb responses to surface variations 3 the majority of the injuries, and the time at risk ,is associated with training (Parkin, 2008; Ramzan and Palmer, 2011). Perkins et al. (2004b) reported the number of events during training periods (Figure 1) that were associated with a rest period, retirement or death. Of a total of 2,652 events reported, 1,594 led to a voluntary rest period (1,234) and retirement (360). The other events reported were musculoskeletal injury (834) leading to a rest period (697), death (19) and retirement (118); respiratory disease (165) leading to a rest period (128) and retirement (37); and miscellaneous (59) leading to rest (38), death (7) and retirement (14). The musculoskeletal injuries (834 cases) involved limbs (807), vertebral column (26) and skull (1). The limb musculoskeletal injuries were localised in the distal forelimb (563), the distal hindlimb (28), the proximal forelimb (26) and the proximal hindlimb (82). Other musculoskeletal injuries were not included in the subcategories as they involved more than one region, or the region affected was not reported. Among all the events reported during both training and racing, the most common type was musculoskeletal injury of the distal forelimb (Bolwell et al., 2017b; Parkin, 2008; Perkins et al., 2004b), as illustrated in Figure 1. The soft tissue structures most frequently affected by musculoskeletal injury are the superficial digital flexor tendon and the suspensory ligament and less frequently the deep digital flexor tendon, its check ligament and the sesamoidean ligaments (Hill, 2003; Rosanowski et al., Figure 1 Distribution of the reported events with number of cases (Perkins et al., 2004b) (MSI = Musculoskeletal Injuries) Chapter 1 Literature review 4 2016). Fractures are most often associated with retirement and death. The risk of a particular type of fracture varies with the type of racing. In Thoroughbreds racing at the Hong Kong Jockey Club, catastrophic fractures most frequently involved the proximal sesamoid bones followed by the carpus and proximal phalanx. Non-catastrophic fractures were observed 3.6 times more frequently than the catastrophic fractures and most often involved the carpus and proximal phalanx (Sun et al., 2019). In an earlier study of racing in the UK, Clegg (2011) reported lateral condylar fractures of the third metacarpal bone to be the most common fracture reported for all forms of National Hunt racing, while in flat racing on all-weather surfaces bilateral proximal sesamoid fractures predominated. In flat turf racing, proximal phalanx fractures were the most common fatal fracture (Parkin et al., 2004b). Injuries have a direct effect on costs, through medical or surgical care, time lost from racing as well as public perception (Perkins et al., 2004b). Musculoskeletal injuries are the major cause of involuntary rest days, with approximately 37% of training Thoroughbreds affected and an average rest duration of 70 days (from 1 to 460 days). Most musculoskeletal injuries require a rest period (83.6%), however they sometimes lead to premature retirement (14.1%) or death (2.3%). The musculoskeletal injuries leading most often to retirement or death are fractures, followed by tendon and ligament injuries. The risk of tendon and ligament injuries increases with the age of the horse and its gender, with the incidence rate for males is 2.5 times higher than for females (Perkins et al., 2004b). In order to reduce these losses, it is important to determine the risk factors for these injuries, which can be grouped into three categories: horse level factors, such as age and sex; race level factors, such as track condition and race distance; and management factors, such as racing load and management of previous injuries (Hitchens et al., 2019). According to Perkins et al. (2004a, 2004b), the main risk factors for the broad category of musculoskeletal injuries are: horse age, gender, cumulative exercise intensity, hoof balance, previous injuries, age at first race, number of starts, physical contact between horses during a race, race distance and class, field size, and barrier position. Among all the identified risk factors, many are not modifiable or unrealistic to prevent, as is the case for gender (Parkin, 2008). On the other hand, Hitchens et al. (2019) suggest that preventing older horses from racing, limiting the number of horses in a race, and avoiding harder surfaces would reduce the risk of musculoskeletal injury. In relation to fracture and dorsal metacarpal disease, the risk factors associated with cyclic load and surface have been the most studied. Repair and adaptation processes are important for the health of the connective tissues of the limb. Exposing a horse to continued or extreme load may overwhelm this process, and thus place the horse at risk of more serious musculoskeletal injuries (Perkins et al., 2004b). Exposure of the tissues of the limb to such loads is determined by training and racing Predicting horse limb responses to surface variations 5 management, which can be changed by regulation or voluntary adjustment of training regimens. A balance needs to be found between too little high-speed exercise and too much high-speed exercise. Indeed, if the exposure to cyclic load causes subchondral bone damage, the lack of high-speed exercise in training would not allow the bone to adapt to the loads experienced under racing conditions. Both scenarios increase the risk of fracture during racing (Parkin, 2008). The role of the racetrack surface as a risk factor for musculoskeletal injury has been described in a number of studies. Different racetrack surfaces have different risk profiles for musculoskeletal injury (Parkin, 2008). This means that it could be possible to reduce the risk of injury to horses training or racing on a track by adapting its design or surface condition (Perkins et al., 2004a). The first surface property studied was the surface hardness, which has been positively associated with the prevalence of lameness in a number of studies (Parkin, 2008). When evaluated in a univariate model, the odds ratio of fatal fracture increased incrementally from 1 (heavy-soft going) through 1.5 (good-soft) to 2.2 (good-firm). However, in a multivariate model including the number of runners and the course distance, odds ratio for fatal fracture increased from 1 (heavy-soft going) through 3.8 (good-soft) to 4.1 (good-firm) (Parkin et al., 2004a). Racetrack surface is still a major area of research (Parkin, 2008). However, athletic injury in racehorses is complex and related to multiple factors affecting both performance and health (Perkins et al., 2004a). The relationships between surface properties and risk of injuries have not been completely established and reducing the risk of a single cause of injury could increase the risk of other injuries (Parkin, 2008). To understand the relationships between racetrack surface properties and risk of musculoskeletal injuries, it will be necessary to fully understand how the tissues of the limb respond mechanically to the surface properties, including stiffness and variability. Modelling in biomechanics Biomechanical models have been developed to illustrate the mechanisms allowing animals to interact with the environment. These models can be classified under three main types: conceptual models, physical models and mathematical models (Alexander, 2003). The conceptual models explain a mechanism by using another one that is well understood. For example, the movement of the foot on the ground during human walking has been compared to an egg rolling from one end to the other. The purpose of this type of model is to clarify and understand simply a complex mechanism without mathematical consideration. However, they are generally not realistic (Alexander, 2003). Chapter 1 Literature review 6 The physical models consist of built structures. This type of model is used to demonstrate that a proposed mechanism actually works, check the output of mathematical models, facilitate observations that would be difficult to make on real organism, explain unexpected phenomena, and determine the consequences of changes in structures. Moreover, physical models indicate why a particular structure is better than another (Alexander, 2003). Finally, mathematical models represent a mechanical output with mathematical equations. These models are used for prediction, seeking an optimum and inverse optimization. This category can be subdivided into four sub-groups: the simple models, the more realistic models, the optimization models and the inverse optimization models. The simple models are the best for establishing general principles. They have been used in human, animal and insect locomotion, to explore effects of changing the properties of tendons and muscles or the number of joints, and to explain other phenomena that can be observed, such as the different shapes of bones or the feeding suction of fish. Some problems may require more realistic models, which are often used to explain more complicated movements, such as somersault or to calculate stresses in elaborately shaped bones. If these models are more realistic than simple models, the assumptions made to develop these models prevent their application in some environments or scenarios. For instance, a model studying the ligament strains at a joint will consider the bones as rigid bodies, and although this model can be accurate for the ligaments it will never be possible to use it to study bone fracture processes. Optimization models enable the calculation of the best structure or pattern of movement possible. For example, they have been used to predict the optimum properties of muscle-tendon units to minimize the metabolic energy costs of movements, and in plants to determine the best patterns of branching to minimize the bending moment and maximize light interception. Lastly, the inverse optimization models are used to test hypotheses, generally about the parameters to optimize, by comparing the results of modelling to actual measurements. This type of model has been used in movement studies, with different optimized parameters: acceleration, muscle power or metabolic energy costs. Another application of this method is to predict forces within the muscles. As there are more muscles than degrees of freedom at a joint, the contributory forces to a net joint moment exerted by each muscle cannot be determined. Optimization functions are needed to determine how the load is shared between the different muscles (Alexander, 2003). Different combinations of optimizations have been tested, depending on the aim of the studies. Seireg and Arvikar (1973) proposed different objectives for optimization of load sharing between muscles: the minimization of the forces in the muscles, the minimization of the work done by the muscles, the minimization of the vertical reaction forces at each joint and the minimization of the moments carried by the ligaments at the joints. Other objectives can be created by combinations of those listed. However, it seems that one optimization function may be Predicting horse limb responses to surface variations 7 specific to a case. The mathematically predicted muscle forces in Herzog and Leonard’s (1991) model of the feline tarsal joint did not agree well the experimentally determined actual muscle forces. They highlighted the variation in load sharing between step cycles as a culprit, and identified that these were due to variation in force-velocity characteristics and/or to delays of onset of activation between muscles. They also indicated that changes in load sharing between speeds may be caused by changes in the magnitude of centrally controlled activation, and none of these principles were considered in their theorical model. To conclude, one of the main risk factors for musculoskeletal injury is the racetrack surface. However, the interaction between the surface and the horse’s limb remains unclear, partly due to our current incomplete knowledge of the adaptive mechanisms of the distal limb, which can be studied using a mathematical model. Hoof-track interaction Risk factors A number of studies have reported an association of track surface and injury in both race and sport horses (Oikawa et al., 2000; Robin et al., 2009; Setterbo et al., 2011; Setterbo et al., 2013; Stover, 2003). Indeed, track surface properties affect both the forelimb hoof impact accelerations and the ground reaction forces (Gustas et al., 2006b; Ratzlaff et al., 2005; Rollot et al., 2004; Setterbo et al., 2009; Setterbo et al., 2011). The most important track surface properties implicated in racehorse musculoskeletal injuries are the hardness and consistency of the surface. The consistency is influenced by the homogeneity of material composition, moisture content, compaction and cushion depth (Cheney et al., 1973; Kai et al., 1999; Mahaffey et al., 2013; Oikawa et al., 2000; Peterson and McIlwraith, 2008; Peterson et al., 2010; Ratzlaff et al., 1997; Setterbo et al., 2013). A greater understanding of the material properties of racetracks, and optimising these might reduce the risk of musculoskeletal injury (Setterbo et al., 2009; Symons et al., 2014a; Symons et al., 2016), and maximise horse performance. Surface properties A number of studies have explored the relationship between the track surface (dirt, turf, synthetic, sand) and the rates and type of injury (Arthur, 2010; Hill et al., 1986; Rosanowski et al., 2016). Due to the varying material properties of the different racetrack surfaces and their effects on shock and vibration of the hoof and distal limb, some types of track surface are associated with certain injuries, such as proximal sesamoid bone fractures on all-weather tracks and proximal phalanx fractures on turf (Parkin et al., 2004b). The higher incidence of certain injuries may be associated with the inherent Chapter 1 Literature review 8 material properties of the different track surfaces. However, there is some variation in the type of injury and the magnitude of the risk associated with different track surfaces reported in the literature. This inconsistency may be due to differences in experimental design, analytic approach, injury and case definition and confounding factors (Setterbo et al., 2009). In addition, variation in environmental factors and surface maintenance procedures have a major impact on the relationship between track surface and injuries, and are hard to control (Setterbo et al., 2011). Rapid loading of the hoof during contact has been identified as an injury mechanism (Pratt, 1997) and track surface properties have been shown to affect the nature of the impact shock and vibrations in the distal limb. The impact shock was attenuated, and the vibration amplitude reduced on an all- weather waxed trotting track compared to a crushed sand track (Chateau et al., 2009b). Vertical ground reaction forces at trot and canter are highest at midstance, and deformation of the hoof capsule, compressive forces across the joints and loading of the suspensory tendons and ligaments are highest at or after this time (Johnston and Back, 2006). These loading events are primarily associated with articular cartilage and subchondral bone degeneration (Chateau et al., 2009b; Radin et al., 1973; Serink et al., 1977). To quantify mechanical loading at impact, tools such as impactors have been developed to replicate the ground reaction forces of a horse in terms of load and speed (Clanton et al., 1991; Mahaffey et al., 2013; Oikawa et al., 2000; Peterson et al., 2008; Peterson and McIlwraith, 2008; Pratt, 1985; Ratzlaff et al., 1997; Setterbo et al., 2011). The rate of loading can be described on live horses through the use of ground reaction force measuring devices or of accelerometers on the limbs (Barrey, 1990; Frederick and Henderson, 1970; Kai et al., 2000; Ratzlaff et al., 1990; Roepstorff and Drevemo, 1993). Impactors Impactors seek to replicate the kinetic impact of the hoof with the surface, and measuring the properties at different locations of a same track provide an indication of the consistency of the hoof track interaction. They have been used to describe a positive linear relationship between the impact force measured and the rate of occurrence of lameness on sand racetracks (Cheney et al., 1973). Using an impactor that measured both vertical force and shear force during simulated hoof landing, Peterson and McIlwraith (2008) reported a reduction in peak vertical load of 34% after harrowing of a dirt track used for Thoroughbred racing, but also that variability of the measured load increased by more than the double across the 24 locations tested. The advantages of impactors are to remove the variability between horses, and to avoid the use of live animals in research (Chateau et al., 2009b; Cheney et al., 1973; Peterson et al., 2008; Ratzlaff et al., 1997; Robin et al., 2009). Predicting horse limb responses to surface variations 9 There has been an increase in the sophistication of the impactors used and their ability to represent the hoof-ground interaction. Pratt (1984) tested the ability of a track to absorb kinetic energy from the hoof. However, his method was cumbersome and inadequate to measure the properties of an entire track (Oikawa et al., 2000). Small loads (10kg) from small heights (under 1m) were used to study only the vertical forces, which are not representative of the triplanar equine hoof impact during racing, and thus do not allow complete characterisation of the surface properties for racing conditions (Pratt, 1985). Clanton et al. (1991) used a cone penetrometer in an impactor with a load cell to measure the force applied to penetrate the soil of a Thoroughbred dirt track to carriable depth. Once again, this method did not describe the complexity of the surface properties, nor did they report their vertical impact velocities. The forces measured during penetration of the soil to 3, 6 and 9 cm only averaged 200, 450 and 1,200 N, which are substantially smaller than the typical vertical ground reaction force (Merkens and Schamhardt, 1994) under the forelimb in a walking horse (about 4,000 N). If the loading rates are not similar to those subjected to the hoof at gallop and if the loads applied are much smaller than required to test the soil to the correct depth, so as to replicate the loading patterns observed in a galloping horse, then the characterization of the surface can only be used in relative terms (Peterson et al., 2008). Indeed, Clanton’s (1991) study was only seeking to characterize the relative surface properties at defined locations across the width of the track and at certain points along its length. Oikawa et al. (2000) combined a self-propelled racetrack hardness measurement device to an analysis system to systematically measure track hardness, and sand depth for dirt surfaces, in all locations of a track. The serial measurements for one entire track were performed across intervals of 5 meters. This device solves the cumbersomeness of the characterisation of a whole track by the devices developed previously. Peterson et al. (2008) developed a mobile testing system that was able to load the track at the rate and loads applied at gallop, at first contact and during the first part of the stance phase when the superincumbent weight was transferred to the hoof. This platform allows the system to be positioned anywhere on the track for sampling the surface and was able to detect changes in the track properties caused by inconsistent surface maintenance. The main drawback of the use of impactors is the strain-rate dependence of the racetrack properties (Ratzlaff et al., 1997; Setterbo et al., 2013). Moreover, studies compared either the peak vertical acceleration, which does not include any consideration of the shear strength of the surface, or used Chapter 1 Literature review 10 small loads, making these tests only representative of the acute impact phase of the horse gait (Peterson et al., 2008). Force plates and hoof-mounted devices The ground reaction force is a relevant property in track surface studies to assess the interaction between the hoof and the ground surface (Robin et al., 2009). In particular, analysis of impact of the hoof and limb may be useful when attempting to improve track conditions (Kai et al., 2000). Two methods have been described to measure the ground reaction force: the use of a force plate or a pressure plate (Pratt and O'Connor, 1976; Robin et al., 2009); or the use of hoof-mounted devices like force measuring shoes, accelerometers and strain gauges (Bjorck, 1958; Parsons et al., 2011). Force plates vs hoof-mounted devices The advantages of using a force plate are the ability to measure actual ground reaction forces in three dimensions and the ease of operation. The disadvantages include expense, the inability to record forces exerted during successive strides, difficulties in obtaining simultaneous recordings of forces exerted by more than one limb, in recording forces at faster gaits and in getting horses to step on the plate, which is the main drawback of this method (Frederick and Henderson, 1970; Kai et al., 2000; Parsons et al., 2011; Ratzlaff et al., 1990; Schamhardt et al., 1993). These limitations generally mean this method is not suitable for measurements in field conditions, at high speed (Robin et al., 2009). Pressure plates have been proposed as an alternative solution to force plates. Compared to force plates, they offer the advantages to allow the analysis of simultaneous and consecutive hoof strikes at once and to provide information on the loading of the different portions of the hoof (Oosterlinck et al., 2010b). However, pressure plates cannot simply replace a force plate when high accuracy of force values is needed (Oosterlinck et al., 2010a).If hoof-mounted devices can overcome many of the disadvantages of the use of force plates, such as measuring the ground reaction force over a large number of strides on a wide variety of surfaces (Kai et al., 2000; Parsons et al., 2011; Robin et al., 2009; Roland et al., 2005), they have other disadvantages. Their volume and weight might affect the gait of the horse. Shoes with strain gauge transducers may not be reliable due to the heavy weight of the device, which may alter the horse’s motion pattern. The measures of the vertical ground reaction forces may also be distorted in devices including strain gauge transducers or piezoelectric transducers by the use of preloaded transducers or by the transducers supporting only a portion of the vertical ground reaction forces. In addition, the first hoof-mounted devices did not supply information about acceleration, which is an important component of the kinetic analysis of gait and gives information on shock and vibration during hoof impact on the ground (Barrey et al., 1991; Hjerten and Drevemo, 1994). Predicting horse limb responses to surface variations 11 Development of hoof-mounted devices Given the advantages offered by hoof-mounted devices compared to force plates, a number of studies have focused on their development and improvement. The first hoof-mounted device was developed by Bjorck (1958). He attached strain gauges to a shoe to measure vertical and horizontal forces exerted by draft horses. The patterns of the force-time curves obtained with this device were typical, in shape, of those obtained by other methods, although the shoe was heavy and thick (Kai et al., 2000). Frederick and Henderson (1970) developed and tested a force-sensitive horseshoe that incorporated three preloaded transducers and obtained the vertical ground reaction forces exerted by a horse at different gaits. Barrey (1990) used an instrumented boot to investigate vertical ground reaction forces exerted at four parts of the hoof. Ratzlaff et al. (1987), Ratzlaff et al. (1990) and Ratzlaff et al. (1993) developed two types of light-weight instrumented shoes using piezoelectric transducers, which measured ground reaction forces exerted over the centre of the frog or at three points on the hoof by horses at different gaits. Roepstorff and Drevemo (1993) equipped a light horseshoe with strain gauges at the toe and at each of the quarters. This device was then used by Roepstorff et al. (1994) to analyse the effect of different treadmill constructions on ground reaction forces exerted by trotting horses. Kai et al. (2000) developed a hoof-mounted device composed of two metal plates, two bolts, four load cells and three accelerometers. The forces recorded from the four load cells (medial and lateral heel, medial and lateral toe) were summed to yield an overall vertical force curve which closely resembled, in both amplitude and shape, the pattern of vertical force measured using a force platform reported by other studies for trot and canter (Gustas et al., 2006b; Merkens et al., 1993). Further the data recorded at trot and canter from two measurement sessions a week apart were not significantly different from each other. Gustas et al. (2004) trotted horses across a force platform, while also collecting data from accelerometers mounted on the fore and hind hooves. The accelerometer signals for the first 50 ms after fore and hind hoof contact were temporally similar but vertical deceleration amplitudes were greater in the fore compared to the hind limbs, agreeing with earlier studies and reflecting the mechanical differences between the functions of the fore and hind limbs. Signals from the force platform likewise indicated similar temporal patterns but with greater rate of loading and larger vertical and horizontal braking forces measured under the forelimb. Roland et al. (2005) developed a 3D dynamometric horseshoe weighting 860 g. It was tested on a treadmill and provided force profiles similar to those reported by studies using other devices (hoof-mounted or force plates). More recently, Chateau et al. (2009a) developed a lighter custom-made device (490g), which was tested on Chapter 1 Literature review 12 several types of ground at slow speed, and was judged to be well adapted to compare the ground reaction forces of different surfaces. The device used by Chateau et al. (2009b) was sensitive enough to discriminate between the biomechanical effects of a crushed sand track and an all-weather waxed track. However, they only tested two kinds of tracks and it is unknown if the device is sufficiently sensitive to identify change of properties within a surface due to variation in moisture level or depth. The main disadvantage of the studies using live horses is the variability between horses. For example, Chateau et al. (2009b) observed the pattern of hoof deceleration during landing composed of two peaks and found differences between three horses. For two of them, they observed a delay of the second peak, but not for the third horse and they were unable to explain the reason for this difference. However, when comparing data, consideration is needed on the mechanism in which the device is used as the stiffness of the surface decreases when the angle of impact increases and when the impact velocity decreases (Setterbo et al., 2011; Setterbo et al., 2013). Thus, the setup of the testing device can have a large impact on the absolute properties reported for a surface. Comparison of surface properties Racetrack surface mechanical properties have generally been compared between surface types (turf, sand, dirt, synthetic). Very few recent studies have compared the properties between traditional surfaces, but have focused on how synthetic surfaces differ to generally one of the traditional surfaces. The difficulty of such comparisons is the maintenance of the ground surface. Indeed, harrowing has been reported as significantly affecting the mechanical behaviour of the surface (Tranquille et al., 2015). Within turf tracks, the turf roots and the soil moisture levels are responsible for increased hardness and resistance to shear (Ratzlaff et al., 1997; Zebarth and Sheard, 1985). Epidemiological studies have identified that turf tracks were associated with a lower risk of breakdown compared to dirt surfaces (Mohammed et al., 1991), however, this will vary with the state of the track. The Jockey Club’s Equine Injury Database (http://www.jockeyclub.com/default.asp?section=Resources&area=10, data for 2019) indicates racing fatalities are lowest on synthetic surfaces (0.93 fatal injuries per 1,000 starts on synthetic compared to 1.56 on turf surfaces and 1.60 on dirt) although the interaction between surface type and age of the horse continues to be equivocal. Whereas older horses tend to be more at risk of injury on dirt surfaces, there is no significant difference in fatal injury rate between 2-year old horses and older horses on synthetic surfaces (Larkin, 2011). http://www.jockeyclub.com/default.asp?section=Resources&area=10 Predicting horse limb responses to surface variations 13 Synthetic surfaces Synthetic racetracks are reported to provide improved consistency and safety compared to dirt tracks (Rezendes, 2007). The synthetic surfaces have been described as generally less stiff and softer than dirt surfaces (Setterbo et al., 2011; Setterbo et al., 2013), and to have better shock-absorbing properties than dirt or sand surfaces (Chateau et al., 2009b; Robin et al., 2009; Setterbo et al., 2009; Symons et al., 2014a; Symons et al., 2016). The maximum vertical forces and loading rates are reported to be lower on synthetic than on traditional surfaces (Chateau et al., 2009b; Crevier-Denoix et al., 2009; Robin et al., 2009; Setterbo et al., 2009; Setterbo et al., 2011; Setterbo et al., 2013). For example, Setterbo et al. (2009) reported peak vertical ground reaction forces of 11.5 N.kg-1, 13.8 N.kg-1 and 16.1 N.kg-1 for synthetic, turf and dirt racing surfaces respectively in cantering Thoroughbred horses, and loading rates of 106 N.kg-1.s-1, 193 N.kg-1.s-1 and 111 N.kg-1.s-1 respectively. However, as previously stated, the maintenance and hydration status will affect the surface properties at any measurement time. The same overall pattern has been reported for the ground reaction forces, maximum impact forces, vertical force peak at impact, vertical force at mid-stance and maximum longitudinal braking force (Chateau et al., 2009b; Crevier-Denoix et al., 2013b; Robin et al., 2009; Setterbo et al., 2009; Setterbo et al., 2011; Setterbo et al., 2013). The peak vertical ground reaction force on the synthetic surface was 83% of the peak on a dirt surface and 71% of the peak on a turf surface (Setterbo et al., 2009). The times of occurrence of the maximal longitudinal force during braking and of vertical force at mid- stance were delayed by 24% and 9% respectively on all-weather waxed surfaces compared to turf surfaces (Crevier-Denoix et al., 2013b), and the time of maximal “sink” of the hoof into the surface was likewise delayed by 30%. A dirt surface, while potentially having more resistance to vertical compression (depending on its depth and maintenance state) will usually allow more horizontal sliding of the hoof, compared to a turf surface that will resist this sliding and therefore allow a greater braking effect across a shorter timespan (Pratt, 1997). Another important parameter to characterize the interaction between the hoof and the track surface is the deceleration of the hoof at impact. On an all-weather surface, the vertical hoof velocity before impact was higher than on a turf surface but the acute hoof deceleration at impact was not significantly different between surfaces (Crevier-Denoix et al., 2013b). Chateau et al. (2009b) recorded a shorter braking phase in trotters on a crushed sand surface (29.7 ms) than on an all-weather surface (35.5 ms) and Robin et al. (2009) associated this with a larger amplitude braking force on the crushed sand track (2,923 N) compared to the all-weather surface (2,392 N). Chapter 1 Literature review 14 The impact of the hoof on the ground surface creates vibrations within the hoof, which are subsequently transmitted to other tissues within the legs. In general, hoof vibrations at impact have lower amplitudes on synthetic surfaces than on traditional surfaces (Chateau et al., 2009b; Robin et al., 2009; Setterbo et al., 2009). This is due to the fact that different surfaces have different vibration energy; for example, low frequency vibrations have higher amplitude on a turf surface compared to an all-weather surface and high frequency vibrations have higher amplitude on an all-weather surface than on a turf surface (Crevier-Denoix et al., 2013b). Stride characteristics have also been compared between the different types of surface. Slip and sink distances during braking and at maximal sink have been reported to be larger on an all-weather surface than on turf (Crevier-Denoix et al., 2013b). Horizontal displacement of the heel during slide is smaller on a synthetic surface than on a dirt surface (Symons et al., 2014a). Shorter stride length and higher stride frequency have been observed on all-weather tracks compared to crushed sand (Chateau et al., 2009b; Robin et al., 2009). The maximum fetlock angle and the heel-strike fetlock angles of the hind limb are smaller and the maximum fetlock angle is delayed on a synthetic surface compared to a dirt surface (Symons et al., 2014a). With all these observations, it seems that synthetic surfaces may mitigate the risk of musculoskeletal injuries (Symons et al., 2014a). However, different results may be observed with different environment and management conditions. For example, the differences between dirt and synthetic surfaces increase as the dirt surface is compacted with repeating impacts or increasing impact velocities (Setterbo et al., 2013) and decrease after harrowing the dirt surface (Setterbo et al., 2011). Moreover, these studies can be affected by large inter-horse variability due to small horse sample size (Chateau et al., 2009b; Crevier-Denoix et al., 2009; Robin et al., 2009). The apparent “better properties” of the synthetic surfaces compared to the traditional surfaces explain the substitution of dirt racetracks with synthetic racetracks (Setterbo et al., 2011). In 2006, the California Horse Racing Board declared that all major tracks in the state must install a synthetic track surface by the end of 2007 (Peterson et al., 2010). With these conversions, fatality rate has been reported to be reduced (Arthur, 2010; Setterbo et al., 2009). This, nevertheless, is just an interpretation as horseshoe regulation and pre-race examination practices also changed at this time (Arthur, 2010). In addition, trainers and veterinarians observed longer race times and more non- catastrophic musculoskeletal injuries. This, combined with the difficulty in managing the synthetic racetrack surface led to the reinstallation of some dirt surfaces (Symons et al., 2016). The observation of greater numbers of non-catastrophic injuries may have been related to the change in surfaces, or may simply have been due to changes in data collection, which are more and more robust. Predicting horse limb responses to surface variations 15 Other comparisons Racetrack surfaces are generally graded according to an ordinal scale that relates to the track hardness and these vary according to the track surface type (dirt or turf). The studies that have used this classification to compare different racetrack surfaces have been epidemiologic. Mohammed et al. (1992) reported higher odds of breakdowns on “sloppy” and “good” tracks than on “muddy” and “firm” tracks respectively (odds ratio of 2.8 and 2.3, respectively). Zebarth and Sheard (1985) reported the odds of serious injury associated with “fast” track conditions was 3.5 times that associated with “heavy” track conditions. Harder, drier or faster race surfaces may be associated with higher risks than rain-affected softer or slower race surfaces (Reiser et al., 2000). In addition to comparison of racetrack surfaces by their class (fast, slow…), it is possible to examine their properties (hardness, shear strength, etc.). A number of studies have related the hardness of track surfaces to an increase of incidence of injuries (Cheney et al., 1973; Drevemo and Hjerten, 1991; Drevemo et al., 1994; Pratt, 1984). Other authors have reported that the ability of a racetrack surface to absorb impact shock reduces the number of breakdowns (Kai et al., 1999; Ratzlaff et al., 2005). Clanton et al. (1991) concluded that the high incidence of breakdowns in one area of a racetrack used for Thoroughbred racing was caused, in part, by the change in slope and compaction of the track in this area. The impact intensity and the shearing forces linked to the horizontal deceleration of the hoof are believed to be important factors in the occurrence of lameness (Cheney et al., 1973; Hjerten and Drevemo, 1994). Impact intensity is related to density and composition of the track (Barrey et al., 1991). Both compaction and composition of the track surface dramatically affect hoof impact deceleration (Barrey et al., 1991; Pratt, 1984). Lower compaction and higher percentages of organic matter result in lower impact forces (Ratzlaff et al., 1997). The compaction of the track surface may also vary broadly over different areas of the same track (Clanton et al., 1991; Drevemo and Hjerten, 1991; Drevemo et al., 1994; Pratt, 1984; Ratzlaff et al., 1997). Some studies on surface physical properties have also related these to mechanical properties. For example, an increase in moisture content of the surface leads to a decrease of the variation in the magnitude of vertical forces between successive strides (Ratzlaff et al., 1997). However, the problem of comparing racetrack surfaces by physical properties is the interdependence of some properties. For example, an increased cushion depth reduces the dry density and hardness of the surface, which results in lower peak decelerations of the hoof at impact; and a reduced dry density leads to a reduced frequency and duration of vibrations at hoof impact (Barrey et al., 1991). There are even more complex relationships between properties. Ratzlaff et al. (1997) reported that when the moisture content was increased up to 8%, returned energy and impact resistance decreased and when the Chapter 1 Literature review 16 moisture content was increased from 8% to 14%, energy returned and impact resistance were progressively increasing. Changes in moisture content also affected the hoof-surface forces, but this relationship depended on the speed of the horses; indeed, they observed the lowest forces at 8% moisture content for horses galloping between 14.5 and 15.4 m.s-1 and at 12% moisture content for horses galloping between 15.5 and 16.5 m.s-1. The hoof-surface forces are also affected by changes in the percentage of energy returned and the impact resistance of the track. For horses with a speed between 14.5 and 15.4 m.s-1, the forces exerted increased as energy returned and impact resistance increased whereas for a speed between 15.5 and 16.5 m.s-1, the forces exerted decreased as energy returned and impact resistance increased (Ratzlaff et al., 1997). Best properties? The objective when installing a new racetrack is to ensure it would have the best properties possible, which implies identifying the surface properties that would minimize the incidence of injuries and maximise the performance of the horses. It is then necessary to understand the role of the track surface in equine locomotion. During the stance phase (the period from first impact to the end of break over) large peak decelerations, the highest vertical load, and highest shear loads are applied (Biewener, 2003; Gustas et al., 2006b; Peterson et al., 2008; Radin et al., 1991), which is why this is the most studied period of the gait cycle. Large high-frequency decelerations in both the vertical and cranio-caudal directions in the early stance phase can have detrimental effects on the musculoskeletal system with associated subchondral bone damage and degenerative changes in the joints (Gustas et al., 2004; Gustas et al., 2006b; Lahm et al., 2004; Lahm et al., 2005; Palmer et al., 1996; Parsons et al., 2011; Radin et al., 1973; Radin, 1999; Serink et al., 1977; Wilson et al., 2001). The forces associated with these large amplitude and high-frequency hoof decelerations are transmitted to more proximal musculoskeletal structures through the hoof as shockwaves (Dyrhe-Poulsen et al., 1994; Gustas et al., 2001; Gustas et al., 2004; Gustas et al., 2006b; Hjerten and Drevemo, 1994; Merkens and Schamhardt, 1994; Parsons et al., 2011; Pratt and O'Connor, 1976; Symons et al., 2014a; Willemen et al., 1999; Wilson et al., 2001). A damping effect has been described, with the attenuations occurring distal to the proximal phalanx (Dyrhe-Poulsen et al., 1994) or distal to metacarpus (Gustas et al., 2001; Willemen et al., 1999). These attenuations have been specified in some research; Lanovaz et al. (1998) and Willemen et al. (1999) described the frequency attenuation as being mainly within the soft tissues of the hoof, while the amplitude attenuation seems to be related to the bones and interphalangeal joints at a slow trot (Gustas et al., 2004). Predicting horse limb responses to surface variations 17 During the impact phase, the hoof is moving at high speed downwards and requires the track to decelerate it and this is the role of the cushion of the track when compressed (Peterson et al., 2008). At first contact, the loading of the limb leads to increased friction between hoof and ground. The vertical deceleration curve shows two peaks during this phase; the first low peak is attributed to heel landing and the second peak to complete landing of the hoof (Chateau et al., 2009b). The horizontal deceleration curve is described as a complex of more or less prominent peaks followed by the distinct local minimum coinciding with the first maximum of the loading rates. The onset of the ground reaction force is characterized by this period (Gustas et al., 2004). The ground reaction force is composed of two principal components; the force associated with acute hoof impact and the one associated with loading of the hoof by the superincumbent limb during stance (Gustas et al., 2001; Gustas et al., 2004; Gustas et al., 2006b; Hjerten and Drevemo, 1994; Parsons et al., 2011; Ratzlaff et al., 2005). The braking phase is defined as the period following impact and during which the hoof still undergoes a sliding movement before complete stabilization on the ground (Chateau et al., 2009b). The forward movement of the hoof coincides with a period of fast extension of the fetlock joint and fast flexion of the coffin joint (Back et al., 1995; Gustas et al., 2004; Johnston et al., 1995). The combination of the sliding hoof, the fast moving distal bone segments and the successive increase in load is suggested to be the cause of the coinciding second complex of vertical deceleration peaks at hoof level, which are also measured at the metacarpus (Gustas et al., 2001). During this phase, there is a rapid increase in the longitudinal braking of the hoof, which appears as a single peak. This peak indicates a horizontal velocity change at the hoof, and shows a large variation in amplitude and timing. The next longitudinal hoof braking peak appears at the time of the second distinct increase in the horizontal hoof braking and metacarpal deceleration (Gustas et al., 2001). The end of this phase is characterized by a more gradual longitudinal deceleration of the hoof. The time period of the horizontal braking of the hoof is also an important factor in the attenuation of the impact (Gustas et al., 2001). Gustas et al. (2006b) concluded that the qualities of the ground surface have an effect on the hoof-braking pattern. The role of the surface during this phase is to help the hoof decelerate, which is also affected by other factors such as the horseshoe design (Kane et al., 1996; Peterson et al., 2008). Ratzlaff et al. (2005) identified an inverse relationship between track rebound rate and negative acceleration peaks of all hooves and concluded that any factors reducing deceleration of the hooves will increase stride efficiency by allowing smoother transition from braking to propulsion and therefore may be important in determining the safety of a racing surface. Chapter 1 Literature review 18 In late stance and during breakover, the horizontal load on the surface is completely reversed to provide a propulsive force. Both braking and propulsion phases determine the properties required for the track surface in terms of shear strength; it has to reduce the magnitude of the abrupt deceleration of the hoof during braking and not fail in shear during propulsion (Biewener, 2