Asymmetries of horses walking and trotting on treadmill with and without rider

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2020 The Authors. Equine Veterinary Journal published by John Wiley & Sons Ltd on behalf of EVJ Ltd 1Department of Anatomy, Physiology and Biochemistry, Faculty of Veterinary Medicine and Animal Science, Swedish University of Agricultural Sciences, Uppsala, Sweden 2Department of Large Animal Clinical Sciences, College of Veterinary Medicine, Michigan State University, East Lansing, MI, USA 3Department of Equine Sciences, Faculty of Veterinary Medicine, Utrecht University, Utrecht, the Netherlands 4Equine Department, University of Zurich, Zurich, Switzerland 5Department of Clinical Sciences, Faculty of Veterinary Medicine and Animal Science, Swedish University of Agricultural Sciences, Uppsala, Sweden


| INTRODUC TI ON
In symmetrical gaits, such as walk and trot, kinematic and spatiotemporal variables should be similar on the left and right sides when the horse travels on a straight line and should be mirror images on left and right circles. Movement symmetry is a highly desirable characteristic in sport horses,1 in which the quality described as "straightness" implies left-right symmetry of limb movements, muscular strength and rein contact. The importance of managing left-right (a)symmetry of the horse's posture and movements is reflected by the fact that straightness is listed as the fifth step in the six-step dressage training scale.1 Dressage test scores, studbook performance testing, pre-purchase examinations. etc., more or less implicitly will value straightness and favour symmetry of appearance and performance.
Considerable attention has been given to quantify asymmetries in kinematics2,3 and vertical ground reaction forces (vGRFs)4,5 in lame horses. However, some level of kinematic and kinetic asymmetries is often found in sound, well-performing horses in both walk6,7 and trot8-11 with considerable overlap being reported between lowgrade lameness and normal horses.12 Laterality, defined as a systematic preference to use one side of the body as a consequence of cerebral lateralisation,13 is discussed as one possible cause for locomotor asymmetries in healthy horses, but other possible factors are conformation,14 trimming and shoeing,15 training16 and pathological or age-related changes. Vertical movement asymmetry increased only marginally in sound trotting horses after the addition of lead weights to the saddle and the rider,17 whereas horses became more asymmetrical with a professional rider vs a novice rider of similar bodyweight.18 This suggests the effect of an active rider goes beyond that of the mere physical load. To the authors' best knowledge, no studies have been published yet on influence of the rider on horse symmetry at walk.
The present exploratory analysis investigated asymmetries in trunk vertical movement, vGRF and spatiotemporal variables in horses assessed clinically as being sound at trot. The aim was to compare movement symmetry of high-level dressage horses at walk and trot in an unridden and unrestrained condition vs ridden between the aids in a dressage frame condition, and to seek associations between corresponding asymmetries in the two gaits. The hypotheses were that asymmetries will increase with a rider and that asymmetries will be related between gaits.

| Horses
The subjects were seven high-level dressage horses (1 stallion and 6 geldings; mean ± SD age: 14 ± 4.3 years; height: 1.7 ± 0.07 m; body mass: 609 ± 62 kg) that were judged by an experienced veterinarian (M.A.W.) to be free from lameness, pain or dysfunction of the limbs and back. The horses were habituated over a period of at least 3 days to treadmill locomotion at walk and trot, with and without a rider.
During their stay at the clinic, horses spent 30-45 minutes daily in a horse walker and were always hand-walked 30 minutes before each measurement session. For unridden trials, the horses wore their own snaffle bridle and were guided by a handler at the front of the treadmill. For ridden trials, horses wore the same snaffle bridle and their customary saddle. They were each ridden by their regular riders (seven riders: body mass: 78 ± 17 kg) who had trained and competed with the horses up to the Grand Prix (n = 6) or Intermédiaire (n = 1) level. The study protocol was approved by the Animal Health and Welfare Commission of the canton of Zürich (188/2005).

| Experimental study set-up
Data were collected at walk and trot (each trial lasted 10-15 s of steady-state gait) without (day 1) and with a rider (day 2) at a range of incremental speeds representative of typical dressage schooling paces within each gait on a high-speed treadmill (Mustang 2200, Graber AG). Unridden horses moved without restraint, guided by a handler at the front of the treadmill. Horses were ridden at walk and sitting trot in a dressage frame, that is in a posture that is typical in dressage with the neck raised, the poll as the highest point and the dorsum of the nose on or slightly in front of the vertical, achieved by coordinated aids from the rider's seat, legs and hands.1 A trial was defined as successful when both a dressage judge overseeing the experiment and the rider agreed that the horse performed a correct and steady gait in good posture, and data were recorded successfully.
The treadmill-integrated force measuring system19 provided vGRF, hoof placement locations and timing for all four limbs.
Unridden force data were sampled at 480 Hz and ridden data at 480 Hz in four horses and 420 Hz in three horses (due to technical difficulties with the kinematic registrations).
Spherical, infrared-light reflective markers (19 mm Ø,Qualisys) were glued to the horses' skin overlying the dorsal spinous processes of the sixth thoracic (T6) and third sacral (S3) vertebrae, and to the left wing of the atlas. Markers were tracked by 12 infrared cameras (ProReflex, Qualisys). Unridden kinematic data were sampled at 240 Hz and ridden data at 240 Hz in four horses and 140 Hz in three horses (due to technical difficulties). The coordinate data were captured using proprietary software (QTrack, Qualisys) and exported for further processing to Matlab (Matlab version 2016b, The MathWorks ® Inc.). Kinematic data were stride-split using the synchronised treadmill data. For each stride and limb, first contact and toe-off were determined by the intersection of the linear K E Y W O R D S horse, gait asymmetry, kinematics, kinetics, sidedness, lameness approximation of the initial and terminal slope of the force curve with the zero-baseline.19

| Kinetic, spatiotemporal and kinematic variables
The following variables were extracted from the treadmill propri- Data quality for these markers was checked by plotting each trial.

| Calculation of asymmetry variables
Left-right asymmetries were quantified by subtracting right-side/ step values from left-side/step values. Stance protraction and retraction differences were calculated by subtracting longitudinal hoof positions at first contact and toe-off respectively. Absolute values for the asymmetry variables were used as dependent variables and unridden/ridden condition was used as fixed effect. Square root transformation of the absolute differences yielded good or adequate normality for all analyses. Models were done separately for walk and trot.

| Data analysis
Model II: to evaluate association between walk and trot asymmetry.
Walk and trot data were matched on horse, speed and unridden/ridden condition. Trials with the lowest and highest speed, and the two trials closest to median speed, were selected and paired between walk and trot for each horse. Atlas, T6 and S3 variables were excluded in model II because of differences in gait mechanics at walk and trot. For the remaining asymmetry variables, two models were made, with either walk or trot as the dependent variable. Fixed effects in the model were the corresponding trot or walk asymmetry variable, unridden/ridden condition and their interaction. In the models with walk as independent variables, walk speed was tested, and in the model with trot as independent variables, trot speed was tested. Both first and second force peaks at walk were analysed vs peak force at trot. Variables were most normal in the untransformed format. Table 1 shows data on number of trials per horse, strides per trial and speed. Figure 1 shows direction of left-right differences, with   Table S1 and Data S1.

| Model I
Absolute value asymmetry variables were analysed to study whether horses were more asymmetric in unridden or ridden conditions (

| Model II
To investigate whether asymmetries were correlated between gaits, associations were sought between values for the same left-right TA B L E 2 Least square estimates and standard error of the mean (SE) from models of absolute asymmetry variables (left-right differences) analysed using square-root transformation and with unridden/ridden condition as a classification effect and speed as a continuous variable  Table S1). Significant differences in unridden/ridden conditions are shown in bold (Walds P value).
Abbreviations: F, forelimb; H, hindlimb.  Table S1 includes information on missing values). The interaction between the asymmetry variable and unridden/ridden condition was significant in two trot and six walk models (shaded grey in Table 3).
A significant association between gaits was found for five variables (bolded in

| Influence of a rider on equine movement symmetry
The current design compares conditions representing two everyday practical situations: the horse moving freely (in the field or in hand) vs when ridden between the aids in a dressage frame, that Note: Each model has data from seven horses, totalling 55 observations (data distributions found in Table S1). There are four (in one case three) observations from each horse and condition. Data were speed-paired: trials with the lowest and highest speeds at walk and trot were paired, as well as the two speeds closest to the median. Grey shading implies that the respective interactions with unridden/ridden conditions were significant and significant slopes are bolded (both Walds P value).
Abbreviations: F, forelimb; H, hindlimb.  Previous studies have looked at the effect on horse symmetry of some of these factors in isolation. At trot, but not at walk, loading the horse with ≥25% of bodyweight led to increased gait asymmetry in ponies.17 Symmetry was measured at the sternum, and the results of that study are therefore in agreement with the increased T6 asymmetry we found ( Table 2). In another study, a skilled rider but not a novice rider made the vertical movement (acceleration) of the horse's sacrum at trot more asymmetric compared with unridden condition. 18 Similarly, in the present study using highly experienced riders, the vertical movement of the horse's sacrum (S3) became clearly more asymmetric with a rider at trot (Table 2). Asymmetry in the interface between horse and rider has also been measured but has been shown to be influenced by both tern makes it likely that the rider will also influence horse symmetry at least to some extent, as was found in the current study.

| Magnitudes of rider-related asymmetries vs lameness
The asymmetry magnitudes in Table 2 can be compared with published studies of lameness. For example, hindlimb retraction differences at trot increased from 20 mm (unridden) to 27 mm (ridden).
Weishaupt and co-authors33 found similar differences of 6-53 mm in horses with mild to more severe hindlimb lameness. Differences in minimum and maximum vertical height between left and right steps for the head and croup (and sometimes also the withers) are frequently used to quantify lameness at trot,34 but only limited data are available for these parameters at walk.2 At trot, asymmetries were 6 mm (unridden) vs 12 mm (ridden) for sacrum vertical maximum, 4 mm (unridden) vs 6 mm (ridden) for withers minimum and 9 mm (unridden) vs 11 mm (ridden) for withers maximum. In a study on induced lameness, similar values were found in the horses not only before induction, but also for the increases in asymmetries when lameness was induced.35 In our study, the rider's weight increased peak vertical force asymmetry only at walk ( Table 2). The left-right differences in first forelimb force peak correspond to 3% and 4% of peak force values (which are approximately 6.5 N/kg at walk) respectively in the unridden and ridden conditions. For comparison, reported forelimb force peak differences at trot were 4% and 9% for mild and moderate lameness.36 To summarise, the increases in asymmetry observed with the riders in the current study were small. We believe differences between unridden and ridden conditions to be below the general detection limit of humans.37 However, the increase in asymmetry with a rider may, in some cases, make an asymmetry just large enough to be perceived by skilled clinicians. Changes between unridden and ridden conditions were analysed using absolute values, but few horses/variables changed in asymmetry direction between ridden and unridden conditions ( Figure 1; Figures S1-S4), indicating consistent sidedness patterns on an individual level.

| If not lameness, why more asymmetry with a rider?
The horses in this study were deemed clinically sound, were in active dressage training and were performing to their riders' satisfaction. Together this suggests that the measured asymmetries in these horses were not pain-related, even if a connection to past or present injuries cannot be fully excluded. Low-grade asymmetries are common in sport horses,9 and laterality has been discussed as one of several possible causes.38 Equine motor laterality has been reported in foals and unhandled youngsters14 and seems to increase with age.39 Motor laterality in horses has been associated with preferences for canter strike-off40 and derailment, which is the tendency to cut across a circle rather than following the circumference.39 However, a connection to kinematic left-right step asymmetry has not been investigated.
It is possible that the rider influences the horse in an asymmetric way, since even skilled riders display postural and kinematic asymmetries,41,42 and riders tend to distribute their weight asymmetrically over the left and right tuber ischia.39 Furthermore, riders apparently have difficulty detecting horse asymmetry while riding44 and that the horse becomes more asymmetric with a rider may be a potential health concern, regardless of whether the asymmetry is caused by the rider's weight or the rider's aids. Carefully designed longitudinal studies with enough subjects are needed to evaluate whether this is a health concern. This study suggests variables that could be used to evaluate a horse's symmetry and straightness of movement both ridden and unridden, but further studies are required to confirm any relationship to laterality.

| Walk vs trot
It has previously been shown that walk kinematics predict trot kinematics. For example, stride duration at walk predicted stride duration at trot. 45 This study evaluated whether asymmetries in either walk or trot would predict the presence and direction of asymmetry in the same parameter in the other gait. In most cases, the association between gaits was similar between unridden and ridden conditions. Only 8 of 22 comparisons showed a significant interaction between unridden/ridden conditions (grey in Table 3). There were five significant associations between gaits, but never for the same variable with both walk and trot as dependent variable. Two of the significant slopes were found for the unridden condition. In two cases, the estimates were negative, suggesting reverse associations for forelimb stance protraction (walk as dependent variable) and the first force peak in the hindlimb (trot as dependent variable). For example, in the unridden condition, the model suggests that increased forelimb stance protraction on the left side at trot will be associated with increased forelimb stance protraction on the right side at walk.
In summary, we did not find strong positive associations between walk and trot asymmetries, perhaps because of the different me-

| Strengths and limitations of the study
An important limitation of this study is the low number of horses.
Studies on larger populations including different ages and stages of training are needed to determine the degree of general applicability.
Furthermore, skin markers were used, which are inherently prone to skin displacement errors. However, if the same marker locations are used, valid comparisons can be made between conditions.47 Furthermore, data were collected during treadmill locomotion which is not identical to over ground locomotion.48 Studying horses first in unridden condition and then in ridden condition may introduce systematic bias. In a study of asymmetry in unridden riding horses trotting overground, it was found that between-trial variation was larger on the first of two consecutive days.49 However, in the current study, the latter effect was likely not of major influence as horses were trained on the treadmill, both with and without rider, for several days before the recorded trials. The strength of this study was the large number of asymmetry variables that were analysed including movements in transverse, longitudinal and vertical directions, in combination with vGRF. Nevertheless, the (long-term) biological significance of the detected asymmetries remains to be addressed. The results for poll (atlas) movement asymmetries must be interpreted with considerable caution, in part because data quality was not optimal and therefore a number of trials had to be discarded, and also because the effects of rein tension applied in the ridden condition could conceivably affect head motion (a)symmetry [c.f. 20]. Rider kinematics were not evaluated in this analysis. P value correction for multiple comparisons was not applied, as statistical power was reduced through analysis of trial level data rather than stride level data.

| CON CLUS IONS
Sound, high-level dressage horses walked and trotted slightly more asymmetrically when ridden between the aids in a dressage frame compared with when moving freely in an unrestrained position.

E TH I C A L A N I M A L R E S E A RCH
The study protocol was approved by the Animal Health and Welfare Commission of the canton of Zürich (188/2005).

OWN E R I N FO R M E D CO N S E NT
Horse owners gave verbal consent for the study.

ACK N OWLED G EM ENTS
We thank the owners and the riders for their committed co-operation.

AUTH O R CO NTR I B UTI O N S
The study used data originally collected by L.

CO N FLI C T O F I NTE R E S T S
None of the authors has any financial or personal relationships that could inappropriately influence or bias the content of the paper.