Several variables must be controlled to avoid variability in kinetic data and temporospatial parameters, including velocity and type of locomotion [1, 17, 20, 21], stance time , training and habituation , body size, conformation, and body weight [1, 2, 4, 5]. In addition, most of the kinetic studies have evaluated dogs that were walking or trotting, due to the symmetry and convenience of these types of locomotion [5, 7, 16, 17]. In the present study, the velocity was maintained 0.9–1.1 m/s and the acceleration between−0.2 and 0.2 m/s2 determined by pressure-sensitive walkway. A training program was not performed in the present study. Because the data are more easily collected using a pressure-sensing walkway compared to a single force plate, the measurements are generally obtained after familiarization to pressure-sensing walkway than a training program [5, 16].
The center of gravity in dogs is located next to the forelimbs possibly near the base of the heart, so that in a healthy dog 60 % of the weight is carried by the forelimbs . The body weight distributions reported in a study of healthy dogs walking on a pressure-sensing walkway, were 60.7 and 39.3 % for small dogs and 61.7 and 38.3 % for large dogs, respectively, for the forelimbs and hind limbs, without influence of body weight or size . In the present study, the mean body weight distributions were similar, being 30 % (G1) and 29.7 % (G2) for each forelimb, and 20 % (G1) and 20.3 % (G2) for each hind limb. Thus, the %BWD may be applicable to comparisons in a heterogeneous group of dogs, because regardless of the body weight, body size, and gait types the values are maintained.
Since the velocity was controlled in the present study, the stride frequency was used to calculate the Pearson correlation coefficients. Besides, the stride frequency is an objective variable calculated by the system, and errors that may occur with tape measurements of the limbs are avoided. The Pearson correlation revealed a strong correlation in all temporospatial parameters analyzing all dogs as unique group, more than analyzing each group individually; suggesting that the gait type did not interfered in this correlation.
The Pearson correlation revealed a strong negative correlation between stride frequency and most temporospatial parameters. Therefore, the values of stance time, swing time, gait cycle time, stride length decreased as stride frequency increased. A study comparing small and large dogs walking at their preferred velocity on a pressure-sensing walkway also reported that most of the temporospatial parameters (gait cycle time, stance time and swing time) were lower for small dogs . On the other hand, a strong positive correlation with swing percentage and a highly negative correlation with stance time percentage were found. Thus, as stride frequency increases, the limb spends proportionately less time on the ground and more time off the ground. Conversely, it was reported that in quadrupeds the swing phase diminishes with increased velocity whereas during trotting and galloping the parameter is quite constant .
With respect to the kinetic parameters, the PVF and VI showed, respectively, low correlation and moderate coefficient values indicating a weaker relationship with stride frequency. A previous study using healthy dogs found that PVF was elevated as the velocity increased and decreased as the stance time increased, while VI decreased as the velocity increased and increased as stance time increased . Thus, other factors may influence PVF and VI, and these parameters may not useful in a heterogeneous group of dogs. On the other hand, no significant correlation was observed between stride frequency and the %BWD, suggesting that the latter parameter was not influenced by the stride frequency.
Symmetry or asymmetry indices or symmetry rates have been used to evaluate kinetic data and temporospatial parameters in dogs walking or trotting over a pressure-sensing walkway, aiming to characterize healthy dogs of the same size or different sizes [5, 17], or to distinguish between lame dogs and clinically healthy dogs [16, 18]. This same strategy was employed in the present study in order to assess the validity of SI in heterogeneous group of dogs, but under controlled velocity.
In both groups the SI of all variables showed median values nearly 0 and asymmetry less than 4 % showing no differences between Groups 1 and 2. These data suggest that these indices could be utilized to evaluate the gait in a heterogeneous group of dogs. However, some facts can limit the use SI for comparison between groups.
A major problem with the SI is that precision depends on the relative magnitude of the evaluated variable . If the magnitude of the variable itself is quite small, such as temporal gait variables in trotting dogs, even small differences may result in high value of SI. Probably, these differences are clinically insignificant, or may be resultant of capture artefacts. On the other, SI of the gait cycle time could be used as an indicative of capture artefacts, since at a constant velocity is not expected asymmetry in this variable.
As an example, the gait cycle time of the forelimb in Group 2 showed 2.58 % of asymmetry (third quartile), which represented a difference of approximately 0.04 s of the mean value of this variable (0.44 s). This difference in mean value of stance phase (0.21 s) can result in a SI of 6 %, and if applied in the dog that showed the lower stance phase (0.13 s) the SI will be 9.1 %. This could explain the high variation of temporal variable SI as well as the SI of VI (total force applied overtime) in Group 2.
In addition, PVF and %BW showed equal values of SIs with a median value near 0. However, a high maximum values can be observed in the boxplots, especially in Group 2. The magnitude of the variable could be a reason to the higher variation in Group 2, but other factors such as velocity variations not evident in trials  and no previous training  must be considered.