Reasons for performing study: Saddle fit is well recognised as an important factor for the health and performance of riding horses. However, only few studies have addressed general effects of different saddle construction details within a group of horses.
Objective: To assess the influence of girth strap placement, traditional vs. v-system, and panel flocking material, wool vs. synthetic foam, on the saddle pressure pattern during riding.
Methods: Six horses were ridden by 3 riders in sitting and rising trot and sitting canter. Saddle pressure was measured with 3 different saddle variants: 1) wool flocked panels and traditional girthing (baseline); 2) wool flocked panels and v-system girthing; and 3) foam filled panels and traditional girthing. From the pressure data, a number of descriptive variables were extracted. These were analysed using ANCOVA models with horse, rider, saddle, seat (sitting/rising, trot only) and speed as independent variables.
Results: With foam filled panels stride maximum pressures under the hind part of the saddle increased by 7–12% and the area under the saddle with a stride mean pressure >11 kPa increased by 114 cm2 in trot and 127 cm2 in canter. With v-system girthing, the latter variable also increased, but only by 53 and 38 cm2 in trot and canter, respectively. In addition, stride maximum pressures under the front part of the saddle tended to increase (≤9%).
Conclusions: Both flocking material and girthing have a significant influence on the saddle pressure and should thus be considered in saddle fitting. Wool seems a better flocking material than foam of the type used in the current study. For girthing, traditional placement seems equally good if not better than the v-system. However, further studies are needed to show if these results are valid for a larger population of riding horses.
The fit of the saddle is well recognised as an important factor for the health and performance of riding horses (Harman 1999). Traditionally, evaluation of saddle fit is done subjectively. Judgements therefore depend on the knowledge, experience and preferences of the saddle fitter. Measurement of the pressure between the saddle and back of the horse offers an objective alternative for assessment of saddle fit, and validity and reliability of such measurements have been previously investigated (Jeffcott et al. 1999; de Cocq et al. 2006, 2009a).
While repeated measurements of saddle pressure with the same horse and rider can be very useful in order to find a suitable saddle for an individual horse (Nyikos et al. 2005), such studies are less relevant for general conclusions on the appropriateness of saddle constructions. Such conclusions require studies where single saddle construction details have been altered while other factors are kept constant, with the effects ideally evaluated in a range of different horses in order to allow generalisation of the results to a larger population. To date, this has only been done for tree width (Meschan et al. 2007). While correct tree fit is necessary for correct saddle fit, other aspects, such as gullet width, panel shape, thickness and flocking material and placement of the girth straps (billets), are equally important (Harman 2004). There are few scientific data that can serve as a basis for recommendations on the construction of saddles, even if such information is often requested by horse owners. Given that the topic is frequently debated among saddle fitters and saddle makers alike, there is a clear need for more objective studies on this topic.
The purpose of the current study was to investigate the influence of girth strap placement, traditional vs. v-system, and panel flocking material, wool vs. synthetic foam, on the saddle pressure pattern during riding, controlling for horse and rider. Additionally, we aimed to quantify the variability between horses and between riders and to assess the possibility that saddle effects differ between horses and between riders.
Materials and methods
Six Warmblood horses, mean ± s.d. height 1.67 ± 0.05 m, were ridden by 3 female riders weighing 53–66 kg. Each horse was ridden by all 3 riders. A fourth rider was used for warm-up of the horses prior to the measurements. All riders were final year students from the Equine Studies programme at the National Equestrian Centre, Strömsholm, Sweden.
The horses were ridden with an experimental saddle custom built on a jumping saddle tree1. This saddle had exchangeable panels fastened with screws and 2 sets of girth straps. One set was attached at the traditional location over the bar of the tree. The other set consisted of a cranial strap attached to the point of the tree and a caudal strap with a sliding attachment to a sling fastened over the middle and rear parts of the tree, a so-called v-system (Fig 1). In the current study the following 3 variants were tested: 1) wool flocked panels and traditional girthing (baseline); 2) wool flocked panels and v-system girthing; and 3) foam filled panels and traditional girthing. Two horses were also ridden with a commercial saddle1 with wool flocked panels built on the same tree as the experimental saddle. The 2 saddles were of equal weight, 8.5 kg including stirrups and leathers. The fit of the saddles was assessed by 2 experienced saddle fitters (K.v.P. and Lars Sundin, constructor of the experimental saddle) and judged to be sufficient, but not perfect in all 6 horses. The study had ethical permission from the Ethical Board for Animal Experiments, Uppsala County, Sweden.
Saddle pressure was measured with the Pliance-X System2. The pressure sensor consisted of 2 halves, each with 128 sensing elements in a 16 × 8 (longitudinal × transverse) array. Sensing element size was 2.5 × 3.75 cm. The sampling rate was 50 Hz. Prior to the measurements the pressure sensor was calibrated for pressures up to 64 kPa by use of a designated calibration rig. Sensor positioning and saddling were done in accordance with the description by Geutjens et al. (2008). The 2 sensor halves were placed symmetrically on each side of the horse's back, aligned with the dorsal midline and longitudinally positioned so that the full length of the saddle panels lay well within the sensor. No additional saddle pad was used. Zero baseline was established before saddling and tightening of the girth and care was taken to always tighten the girth to a similar extent. The saddle and pressure sensor were repositioned and the sensor re-zeroed between each rider and saddle variant.
Measurements were taken along the long side of an indoor riding arena. Each horse and rider combination was measured in sitting and rising trot and sitting canter on both left and right hands. In total, 360 measurements were made and 10–15 strides recorded in each measurement. Speed was measured using a stopwatch and 2 markings on the kickboard of the arena. The riding order for the riders was randomised for each horse. Traditional girthing and v-system were measured in a random order for each rider. The foam panels were measured at a separate session, to avoid fatigue failure of the panel screw attachments due to multiple panel exchanges.
Pressure raw data were exported to Matlab3 for further processing. Previous research has found a close timing relationship between saddle pressure (or force) grand minima and hindlimb ground contacts in both trot and canter (Fruehwirth et al. 2004). Therefore, these minima were used for a stride split of the raw data in the present study. The frame sum of pressure over time was calculated for each measurement. Frame indices for grand minima (Fig 2) were determined and the indices for every other minimum for trot measurements and all minima for canter measurements were collected and formulated into a stride index. Mean stride duration and frequency were calculated. The raw output from each sensing element was then low pass filtered with a cut-off frequency of 6 × the stride frequency. After filtration, the raw data were normalised in space to a pixel size of 1 × 1 cm and in time to 0–100% of stride cycle duration (101 points).
In order to reduce noise, a mask or template defining the area of interest was created. The pixel selection criteria were the following: 1) stride mean pressure >2 kPa, as Pliance's lower accuracy limit is ∼2 kPa and 2) stride mean between-stride coefficient of variance (CoV) ≤50% (>50% being rare for pixels with a stride mean pressure >2 kPa). In order to include equivalent areas on the left and right sides, the right mask was reflected and summed upon the left mask. A copy of the summation mask was subsequently reflected back to the right side to produce a final left-right symmetric mask. All pixels outside the final mask were set to zero.
From the masked and normalised strides, a mean stride pressure sequence was determined and the total force, i.e. sum of pressure × pressurised area (Meschan et al. 2007), was calculated for each normalised frame. Additionally, the frames corresponding to the total force maxima indicated in Figure 2 were extracted from each normalised stride and averaged over available strides, giving 2 mean pressure pictures representing total force maxima for each measurement.
In order to compare the different saddle variants, the present study employed variables previously used or suggested as relevant for the assessment of saddle fit: stride maximal total force (Meschan et al. 2007), area with measurement minimum pressure >4.67 kPa/35 mmHg (Harman 1994, 1997), area with stride maximum pressure >30 kPa and area with stride mean pressure >11 kPa (Werner et al. 2002; Nyikos et al. 2005) and stride peak pressure in each quadrant (inner front, outer front, inner hind, outer hind) of the pressurised area (Mönkemöller et al. 2005; Nyikos et al. 2005; Meschan et al. 2007). As an objective alternative to subjective evaluation of pressure evenness (Werner et al. 2002), the mean pressure gradient at total force maxima was used. The mean pressure pictures representing total force maxima were each subjected to gradient transformation using the following formula:
Following gradient transformation, a mean was determined over all nonzero pixels. This mean, however, excluded pixels with a stride mean ≤2 kPa or a stride mean between-stride CoV >50% and pixels adjacent to such pixels. The reason for this was that edge enhancing operations, such as gradient transformation, tend to enhance noise (Gonzalez and Woods 2008). In sitting trot the diagonals could not be identified as left or right. Therefore, the average of the mean pressure gradient at each diagonal's maximum was used for sitting trot.
The mixed procedure in SAS4 was used to create multivariable models with horse, rider, saddle, seat (sitting/rising, trot only) and speed (modelled as a linear effect) as independent variables and the different saddle pressure variables (including stride duration) as the dependent variable. Trot and canter were analysed separately. For trot, the interaction between saddle and seat was also included. Horse and rider were modelled as random factors and interactions between these and saddle were included if this improved the Akaike information criteria (AIC) for the model. Speed was expressed as the difference from gait mean speed, and as a result the intercept represents mean speed. Speed was forced into the model as a linear relationship is known to exist between speed and saddle force (Bogisch et al. 2008). Bilateral variables were labelled as inner and outer rather than left and right and measurements on left and right hands where regarded as repetitions. For fixed effects, the P value limit for inclusion in the final model was <0.05. If the P value for saddle was <0.1 saddle was added to the final model in order to assess tendencies (occurred only once). Model parameter estimates were used to find significant differences between intercept (baseline), i.e. riding at gait mean speed with wool flocked panels and traditional girthing with the rider sitting in the saddle, and other experimental conditions. For quadrant peak pressures, the estimates were expressed as a percentage of intercept, in order to facilitate interpretation. For the other dependent variables, least square (LS) means were used to quantify variable values for the different conditions (LS mean for baseline conditions corresponds to intercept). The variance estimate for each random factor was noted and expressed as a percentage of the total random variation in the model.
Because of the linear relationship that exists between the saddle pressure (or force) sum and the weight of the rider and saddle (Jeffcott et al. 1999; de Cocq et al. 2006, 2009a), forces that were measured in this current study were divided by the weight of rider and saddle in Newtons. Weight normalisation has also been used for peak pressures during mounting (Geutjens et al. 2008) and peak pressures measured in the current study were therefore expressed as Pa/N rider and saddle weight. Gradients and areas were not normalised because weight normalisation failed to reduce between-rider variability and because pressure recommendations in the literature are given without regards to rider weight (Werner et al. 2002; Nyikos et al. 2005), respectively.
Trotting speed was (mean ± s.d.) 3.95 ± 0.22 m/s and was not significantly different between sitting and rising trot. Speed in canter was 4.63 ± 0.18 m/s. Stride duration was (LS mean ± s.e.) 0.77 ± 0.015 s in trot and 0.61 ± 0.010 s in canter and was not significantly affected by either saddle variant or seat (sitting/rising). Stride maximum total force was (intercept ± s.e.) 2.7 ± 0.09 ×‘rider and saddle weight’ in trot and 2.8 ± 0.15 in canter. There was no significant difference between saddle variants, but rising trot had a significant effect, as reported below.
The effects of the different saddle variants on quadrant peak pressure are illustrated in Figure 3a and b, and effects on the different area variables and on mean pressure gradients at total force maxima are illustrated in Figure 4a and b, respectively.
When the saddle panels were exchanged for foam filled panels stride maximum pressures under the hind part of the saddle became significantly higher and mean pressure gradients at total force maxima also increased significantly, in both trot and canter. Further, all area variables increased significantly in both gaits, except for the area with stride maximum pressure >30 kPa during rising trot.
When girthing was changed to the v-system, stride maximum pressures increased under the front part of the saddle in trot and canter, but the mean pressure gradient increased only at the total force maximum of the rising diagonal in rising trot. Of the area variables, only the area with a stride mean pressure >11 kPa increased significantly.
With stitched panels, stride maximum pressures tended to increase under the hind part of the saddle and decrease under the front part of the saddle. In canter this was more pronounced on the leading side compared to trailing side. The mean pressure gradients at total force maxima were not significantly affected and neither were any of the area variables.
Compared to sitting, rising trot resulted in a marked increase in stride maximum pressure under the outer front quadrant of the saddle. For all other quadrants stride maximum pressure decreased significantly. Stride maximum total force decreased by 0.46 ± 0.02 ×‘rider and saddle weight’ (estimate ± s.e., P<0.001) and the mean pressure gradients at total force maxima also decreased significantly (Fig 4b). However, the area under the saddle with a stride mean pressure >11 kPa and with a measurement minimum pressure >4.67 kPa became larger.
There was a significant speed effect for the following variables: maximum total force and peak pressure in the outer hind quadrant in both trot and canter, mean pressure gradient at the total force maximum of the sitting diagonal in rising trot (rising diagonal P = 0.099) and mean pressure gradient at the hindlimb stance total force maximum and area with stride maximum pressure >30 kPa in canter. The slope ranged from 0.6–2.4% per 0.1 m/s, except for the area with a stride maximum pressure >30 kPa in canter, which increased by 8% per 0.1 m/s.
Sources of random variation in the models
For most variables, variation between horses constituted 20–50% of the total random variation in both trot and canter models. Higher values, 59–95%, were found for stride duration and maximum total force, and lower values, <20%, were found for peak pressure in the outer hind quadrant in trot and the area with a stride maximum pressure >30 kPa in both gaits. Adding the interaction between horse and saddle resulted in an improvement of the AIC-value for all variable models in either one or both gaits, except for peak pressure in the outer front quadrant. The horse-saddle interaction contributed to the total random variation by a value between 1 and 29%. The highest values were found for the area with a measurement minimum pressure >4.67 kPa.
The rider's contribution to the total random variation ranged from 0.6–63%. Mean gradients at total force maxima and the area with a stride maximum pressure >30 kPa had the highest values, while stride duration and the area with a measurement minimum pressure >4.67 kPa had values ≤3% in both gaits. Adding the interaction between rider and saddle improved the AIC-value of the models for peak pressure in the outer hind quadrant and mean pressure gradients at total force maxima in trot and for maximum total force in both gaits. This interaction accounted for 3–6% of the total random variation for these variables.
In the current study, the pressure recordings of the saddles were found to be in agreement with the subjective judgements of the saddle fit. By comparing the results to previously published criteria (Harman 1997; Werner et al. 2002; Nyikos et al. 2005) it can be concluded that saddle fit was not optimal in all horses, yet within acceptable limits. Further, while flocking material and girthing had a significant influence on several variables, the magnitudes of these effects were not dramatic in relation to changes that occur during normal riding, such as changing from sitting to rising trot. The differences observed between saddle variants in the current study should thus not be interpreted as differences between perfect and ill-fitting saddles. Rather, the results show that flocking material and girthing can influence the pressure under moderately fitting saddles and that such effects can be objectively quantified.
The experimental saddle used in the current study had panels that were fastened with screws rather than stitches to allow the panels to be exchanged. This construction may have caused some bias to the results and thus measurements were also made with a saddle built on the same tree as the experimental saddle but with normal stitched panels. With the stitched saddle peak pressures were lower under the front part and higher under the hind part of the saddle. The screw attachments for the panels made it difficult to pack the flocking material as tightly in front as with stitched panels and this may have influenced the cranio-caudal balance of the saddle. Increased peak pressures under the hind part of the saddle were also observed with foam panels, but were in this case accompanied by higher mean pressure gradients at total force maxima. Foam panels thus seem to have caused more pronounced local peaks, rather than just a shift in the cranio-caudal balance of the saddle. Increased peak pressures were also observed with v-system girthing, but under the front rather than the hind part of the saddle. For the v-system a short saddle girth was used (Fig 1), which may have influenced the degree of tightening of the girth, in turn possibly influencing the saddle pressure to cause this effect. However, with the v-system, the cranial girth strap was placed further forward, which could have pulled the saddle down more in front by leverage effect, despite similar girth tension. The effects of the v-system may also have been more pronounced as a result of the experimental saddle having some tendency towards bridge formation or 4 point pressure in several of the horses. Therefore, the v-system mainly seems to cause a shift in the cranio-caudal balance of the saddle. However, in the outer front quadrant, pressure increased both with the v-system and during rising trot and when these 2 were combined the mean pressure gradient at the total force maximum of the rising diagonal increased significantly.
Objective evaluation of saddle fit is, of course, a question of which variables to analyse, how to interpret values and how to prioritise between different variables. In 2 of the first publications on saddle pressure, 35 mmHg (4.67 kPa) was discussed as a possible limit (Harman 1994, 1997) due to the fact that pressures above this limit are known to significantly reduce skin perfusion in man (Holloway et al. 1976; Chang and Seireg 1999). Mean and maximum pressures when riding with an English saddle and an average size rider are typically considerably higher, but it may make a difference that the pressure is relieved below this limit at least once during each measurement of 10 s. Area with a measurement minimum pressure >4.67 kPa was therefore compared between saddle variants in this current study. Nyikos et al. (2005) reported that for sitting trot maximum pressures >34.5 kPa under the front part of the saddle and >31 kPa under the hind part of the saddle as well as mean pressures >13.2 kPa and >10.0 kPa, respectively, correlated to back pain. In the current study these limits were generalised to >30 kPa and >11 kPa and areas with these undesirable properties were compared between saddle variants. From a theoretical standpoint it should be asked what is more harmful - high maximum or high minimum pressures. However, in the current study maximum and minimum pressures co-varied to affect mean pressure, which showed the largest differences between saddle variants (Fig 4a). For all areas, in particular the mean pressure, the effects of flocking material as well as girthing were also clearly larger compared to the effect of rising vs. sitting trot. The ideal variable for objective assessment of saddle fit should be both sensitive and specific and in the current study ‘area with stride mean pressure >11 kPa’ seemed to fulfil these criteria best. In a recent study comparing saddle pressure during riding between horses with saddle sores, horses that were not sweating in focal spots under the saddle and horses without any back or saddle problems arrived at a similar conclusion: mean pressures differed more clearly between the groups compared to maximum pressures (von Peinen et al. 2010). Thus, evaluation of saddle fit seems best performed when the entire stride and also the entire pressurised area is considered. In future studies it would therefore be interesting to seek novel variables that better represent the pressure dynamics throughout the stride cycle. Principle component analysis may be a way forward (Witte et al. 2009).
For many of the variables analysed in the current study, several factors other than saddle, such as horse, rider, seat and speed, were found to have a significant influence. Possible effects of girth type and girth tension were not investigated but this may be advocated in future studies as it has been shown that the force applied by the saddle is significantly higher with a tightened girth compared to a loose girth (de Cocq et al. 2006, 2009a). Rising at the trot had a significant effect on most variables in the current study. The observations are in agreement with previous reports (Peham et al. 2009; de Cocq et al. 2009b). Speed had a significant effect on several variables and, due to the narrow speed range studied, actual speed influence on more variables cannot be excluded. Influence from speed has been confirmed by others (Bogisch et al. 2008) and thus speed seems an important factor to control in saddle pressure studies. Almost all variables showed considerable variation between horses. For several variables there was an interaction between horse and saddle, with saddle effects differing between horses. This seems reasonable, considering that the initial saddle fit was not identical in all horses. For several variables the rider was also a prominent source of variation. For some variables there was additionally an interaction between rider and saddle, despite the fact that the riders were all upper medium level and had 2 full years of equal training and instruction prior to the study (within the Equine Studies programme). This is important, because it suggests that the number of riders used for each horse may constitute a limiting factor for generalisation of results to a larger population. In future studies on saddle fit it therefore seems advisable to use several riders for each horse, to allow quantification of between-rider variability and detection of significant interactions between rider and saddle fit.
In conclusion, flocking material and girthing have a significant influence on saddle fit and should be considered when a saddle is fitted to a horse. In the selection of flocking material, wool seems a better option than foam, at least the type used for the current study. Studies on wheelchair cushions and mattresses in man (Apatsidis et al. 2002;Bain et al. 2003) have shown that different types of foam have different properties, so it cannot be stated that wool is always preferable to synthetic materials. For girthing, traditional placement seems equally good if not better than the v-system. However, before any general conclusions can be drawn, girth strap placement needs to be evaluated for a range of different saddles. Our results show that it is possible to quantify and characterise the effects of different saddle construction details by objective measures. Further studies are needed to show if saddle pressure effects of the kind and magnitude observed within the current study are biologically significant in terms of horse health and performance.
The authors would like to thank certified saddle maker Lars Sundin for providing the experimental saddle and the riders for their participation. This study was funded by grants from Stiftelsen Svensk Hästforskning and Ulla Håkansons Stiftelse.
Conflicts of interest
The authors have declared no potential conflicts.
1 Morris and Nolan Saddlemakers, Walsall, West Midlands, UK.
2 Novel GmbH, Munich, Germany.
3 The Math Works Inc., Natick, Massachusetts, USA.