This study demonstrated that compared to the normal group, the orthopaedic group had increased body weight distribution and increased stride velocity in the forelimbs, while the neurologic group had increased stance time in the forelimbs without significant changes in body weight distribution or stride velocity. These results support our hypothesis that PSW could detect distinctive changes in kinetic and spatiotemporal gait parameters in dogs with bilateral hindlimb lameness and in dogs with hindlimb ataxia. Additionally, the results also indicate that dogs with bilateral hindlimb lameness and hindlimb ataxia compensate for their gait abnormalities through different mechanisms; thus, PSW analysis may help in the differentiation of neurologic versus orthopaedic hind limb disease.
For the kinetic parameters, the orthopaedic group had significantly greater %BWD and VID in the forelimb, as well as greater forelimb:hindlimb SI maximum force, than did the normal group. This indicates that dogs with bilateral hindlimb orthopaedic disease compensate by shifting their weight to their forelimbs; dogs with neurologic disease did not demonstrate this effect. In a healthy dog, weight is approximately distributed 60% to the forelimbs and 40% to the hindlimbs, regardless of the body weight or size [27–30]. The median body weight distributions of the normal group in the current study were 28.3% (range: 25.9–32.9%) and 29.00% (25.0-33.1%) for the forelimbs and 21.6% (17.4–24.1%) and 20.65% (17.2–24.8%) for the left and right hindlimbs, respectively, which are comparable to the previously reported normal values for dogs and confirmed that the normal group was valid as a control.
Although “compensatory cranial weight shift” in dogs with bilateral hindlimb orthopaedic diseases is commonly described in the clinical setting, objective data based on comparisons with a control group are sparse [31]. Multiple studies evaluating the weight distribution of dogs with various naturally occurring or experimentally induced lameness have shown that dogs with lameness tend to redistribute their weight to the non-affected limbs, mainly to the contralateral and diagonal limbs, thus exhibiting more side-to-side compensation rather than a caudal-to-cranial shift [23, 28, 32–34]. However, a recent study utilizing a stance analyser to evaluate weight-bearing compensation in police-working dogs with bilateral hip osteoarthritis revealed that affected dogs had weight shift to the thoracic limbs, corroborating the common clinical description of “cranial weight shift” in bilaterally affected dogs [35]. The results from the current study are not only in line with these findings but also demonstrate statistical differences in the kinetic parameters in comparison to those of a normal control group to provide further objective parameters in support of the commonly accepted cranial weight shift phenomenon in dogs with bilateral hindlimb lameness.
In contrast to the orthopaedic group, the neurologic group showed similar body weight distribution in all limbs compared to the normal group, indicating that compensatory mechanisms may be different in dogs with neurologic disease. Changes in kinetic parameters in dogs with various neurologic diseases have been investigated in previous studies [36–39]. One study evaluated kinetic gait parameters in Doberman Pinchers with cervical spondylomyelopathy and found that there was no significant difference in the %BWD of the forelimb or the hindlimb compared to clinically normal dogs [36]. Another study revealed no significant differences in the PVF between the hindlimbs of normal Dachshunds and Dachshunds that underwent hemilaminectomy for T3-L3 intervertebral disc disease [39]. This study also showed that the post-hemilaminectomy group had greater PVF on the more affected limb, which was contrary to findings in dogs affected by orthopaedic disease where lower PVF is expected in the limb with impaired function [5, 6, 23, 39, 40]. Along with the results from the current study, these findings support that dogs with ataxia may not necessarily redistribute their body weight to compensate for their uncoordinated gait. An alternative explanation is that these dogs could not control the force exerted by the limbs due to loss of input from the upper motor neuron and were “unable to” consciously compensate for the instability in the hindlimbs.
Interestingly, in another study that evaluated gait parameters in dogs with ataxia due to T3-L3 myelopathy, the affected dogs had greater PVFs in the forelimbs than did the normal dogs [21]. The authors of that study postulated that dogs with T3-L3 neurologic disease tend to shift their weight to the forelimbs as a result of hindlimb ataxia and instability [21]. This finding is in contrast with our data. The disparity in the observations between the studies may be due to several factors. In the aforementioned study, Dachshunds were overrepresented, composing 70% of the neurologic group in that study, while the neurologic group in our study included a more diverse population of dogs of various sizes, breeds, and conformations, with only 3 of 11 dogs being chondrodystrophic breed. Although the influences of the variation in the body weight and sizes of dogs can be avoided by the use of kinetic values normalized to the body weight, one study has suggested that there can be significant differences in the fully normalized ground reaction force and impulse distribution in the forelimb versus hindlimb between various dog breeds [26, 41]. Therefore, the overrepresentation of a certain breed or body conformation may have led to different results in the distribution of body weight [41]. Additionally, the difference in the velocity range of the subjects and the method of normalization of this value may have led to different outcomes. It is known that kinetic gait parameters can be dependent on the velocity and acceleration of patients [42–44]. To avoid this influence, previous investigators have suggested maintaining the variables within a range during gait analysis [42–44]. In our study, all dogs were walked at velocities between 0.8 m/s and 1.4 m/s, in accordance with protocols validated in previous studies [42–44]. In contrast, the neurologic group in that study first walked at their preferred pace, which initially led to a significantly lower velocity than that of the clinically normal group [21]. Although the difference was not significant after the velocities of the groups were adjusted for the heights of the subjects by regression analysis, these overall differences in the data collection protocol and processing method may have contributed to the discrepancies in the gait variables in different studies.
In the present study, when comparing individual limbs, the %BWD and VID of the LF, RF, and RH achieved a significant difference between the normal group and the orthopaedic group, but the values for the left hindlimbs were closely approaching but did not achieve a significant difference (p = 0.051). This is likely due to this cohort of dogs being most lame on the right hindlimbs, as demonstrated by the negative maximum force SI of the right: left hindlimbs of the group. Combined with the relatively small group size, this may have led to a type II error. It is possible that if the side of dominant lameness was evenly distributed, the kinetic variables in the LH may have reached statistical significance.
In addition to the differences in the redistribution of body weight and kinetic parameters, the neurologic group and the orthopaedic group demonstrated distinctive changes in spatiotemporal gait parameters compared to the clinically normal group. One of the notable changes in the neurologic group was the increase in the stance time of the forelimbs. When compared within a group, all groups had positive forelimb: hindlimb SI of the stance time, indicating that dogs in our study had a longer stance phase in the forelimbs than in the hindlimbs, regardless of their group. This finding is in agreement with other studies that demonstrated longer stance times on the forelimbs than on the hindlimbs in dogs at a walk [45, 46]. When comparing between groups, the forelimb: hindlimb SI stance time of the neurologic group was significantly greater than that of the normal group, which was not observed in the orthopaedic group. Instead, the orthopaedic group had relatively greater stride velocity in the forelimbs than the hindlimbs compared to the normal dogs. These distinctive changes in spatiotemporal gait parameters in the orthopaedic and neurologic groups are likely reflective of different compensatory mechanisms for hindlimb instability: dogs in the neurologic group may stabilize their unsteady gait in the hindlimbs by increasing the time that the forelimbs are in contact with the ground, while dogs in the orthopaedic group may take quicker strides in the forelimbs to afford the increased weight distribution.
In our study, we found no significant differences in forelimb:hindlimb SI of the stride time or stride length among the groups. The neurologic group was the only group with negative mean values, as well as the lowest median forelimb: hindlimb SI values for stride time and stride length, but the differences between the neurologic group and the other groups did not reach statistical significance. Although our study did not find statistically corroborating results, a published study reported decreased stride time and stride length in the forelimbs of dogs with T3-L3 myelopathy and hindlimb ataxia [22]. It has been shown that dogs with neurologic disease have greater variances in the spatiotemporal gait parameters compared to clinically normal dogs[21]. A larger sample size with the variables normalized to the height of the subjects may be required to detect statistically significant differences. In addition, the severity and chronicity of disease in the neurologic group were not specified in the current study. These factors may have affected the degree or pattern of gait compensation. Future studies focusing on a specific disease process with a similar chronicity and neurologic grading may lead to different results.
The overall changes observed in the neurologic group included an increase in forelimb stance time and a relatively lower stride time and length. These findings are similar to those of a previously reported gait analysis study in rats with experimentally induced spinal cord injuries, in which the rats had increased stance time and decreased stride length in the forelimbs [47]. On the other hand, the orthopaedic group had increased body weight distribution and stride velocity in the forelimbs compared to those on the hindlimbs, without any increase in stance time. Thus, this can be interpreted as dogs with bilateral hindlimb lameness compensate by shifting body weight cranially, which subsequently increases the stride velocity of the forelimb to support weight shift during each gait cycle. Based on these observations, gait parameters such as %BWD, VID, forelimb:hindlimb SI of stance time, and stride velocity are worthy of further investigation to evaluate their potential to discriminate between the two conditions. A further study that determines the cut-off value for each variable and receiver operating characteristic (ROC) curve may assess the accuracy of the subset of these parameters in discriminating dogs with bilateral hindlimb lameness and ataxia based on the data obtained by the PSW.
The main limitation of the study was the relatively small sample size, which may have led to a type II error. This was demonstrated by the apparent sidedness observed in the orthopaedic group, even though all dogs within the group were confirmed to have bilateral orthopaedic diseases and decreased weight-bearing in both hindlimbs, as confirmed by gait analysis. The small sample size may also have affected the gait parameters for the neurologic group, especially when they had the greatest variability within the dataset, with greater ranges in many of the analysed gait parameters. While all dogs included in the neurologic group were classified as grade 2 (ambulatory paraparesis) based on the modified thoracolumbar spinal cord injury scale, the clinical status of individuals varied greatly from mild to severe ataxia. This inherent variability of the data in subjects with neurologic disease has been noted in both humans and dogs, and the use of the coefficient of variation of variables, rather than individual mean values, has been advocated for these patients[22, 48, 49]. Further studies including larger sample sizes and three-way comparisons of the coefficient of variation of the neurologic group, the orthopaedic group, and the normal group may reveal further discriminating gait parameters and patterns among the groups. Larger sample sizes will also help to build reliable predictive models based on the discriminating gait parameters found in the current study, which can be used to evaluate the reliability of the PSW as a diagnostic tool.
Our study included a heterogeneous group of dogs of various breeds, weights, and body conformations. Although this is a more accurate and practical presentation of clinical settings, one study suggested that various breeds, body conformations, heights, and weights of dogs can affect spatiotemporal and kinetic gait parameters[41]. There were no significant differences in age or sex among the groups in the current study, but the mean weight of the orthopaedic group was greater than that of the other groups. This was reflective of the fact that many of the dogs within the group were large breeds that presented with bilateral cruciate ligament diseases. While authors are aware of this potential source of bias, it also reflects patient demographics encountered in a clinical setting. Additionally, a recent study showed that the %BWD and most SI values have low variability in a heterogeneous dog group [26]. The SI value also has a benefit in that it eliminates interpatient variability, as the patient serves as its own control. Thus, we focused on the comparison of the normalized kinetic variables and SI values to minimize the potential influence from the breeds and confirmation by comparing the changes in variables within the patient.
In conclusion, the orthopaedic and neurologic groups exhibited distinctive changes in spatiotemporal and kinetic gait parameters compared to those of normal dogs. The findings also suggest that dogs with hindlimb gait abnormalities may have different compensatory mechanisms for their gait deficiencies depending on whether their conditions are orthopaedic or neurologic in origin. Compared to those in the normal group, significant differences were found in gait parameters such as the %BWD and forelimb: hindlimb SI values of the stride velocity of the orthopaedic group and in the SI stance time of the neurologic group. In the future, a larger-scale study may help to determine the optimal cut-off value and build predictive models based on the discriminating gait parameters found in the current study. This may further support the utility of the PSW in differentiating between two conditions that are often difficult to determine clinically.