Objective gait analysis is not commonly used in small animal clinical practice because it can be time consuming and requires expensive instrumentation. However, objective measurement and evaluation of gait variables are useful in distinguishing normal from abnormal gaits and for assessment of long-term outcomes for surgical procedures and nonsurgical interventions.1–3 To date, force plate analysis has received the most attention in both human and veterinary medicine,4–9 and it is widely considered to be the gold standard of objective gait assessment. In quadrupeds, the sequential use of 2 (or more) force plates has been shown to improve accuracy and efficiency of data collection.10
Pressure-sensitive walkways were introduced as a time-efficient alternative to the use of force plates. These walkways have the important advantage of providing both temporal and spatial information on the kinetics of all 4 limbs during the gait cycle.11 With a walkway of the appropriate length and width, it should be possible to collect data for 1 or 2 complete gait cycles from dogs of different sizes. Evaluations of various walkway systems, ranging from pressure-sensitive plates to full walkways, have been published.10–12 Determination of step length, stance phase, and swing phase allows walking trials to be divided into individual gait cycles, each consisting of a stance phase and swing phase. This permits simple comparisons of forces and paw contact areas across multiple strides and walking trials.
Pressure-sensitive walkways have some important limitations, the most important of which is that it is not possible to have precise control of the velocity at which the dog moves over the walkway. It has been shown that velocity is a key determinant of temporospatial variables for dogs,13 and thus, the inability to control velocity contributes to variability of data collected from pressure-sensitive systems that record over a fixed-length walkway. Walkways also require a large amount of floor space in the gait analysis laboratory or clinic (typically at least twice the length of the walkway is needed to allow animals time to reach a stable walk or trot before reaching the walkway and then slow down after stepping off of it). Finally, pressure-sensitive instrumentation can be very expensive, even relative to force plates, and it can easily be damaged if the sensors are not handled carefully.
Instrumented treadmills were introduced in the late 1980s as a space-efficient and cost-effective alternative to instrumented walkways.14,15 Instrumented treadmills are available that make use of force sensors16,17 or pressure sensors.18,19 Although introduced initially for the human market, instrumented treadmills are becoming more common in the veterinary market. Brebner et al20 reported on the use of a forcesensitive treadmill for assessment of dogs with and without lameness, and Assaf et al21 recently compared gait analysis results measured with a pressure-sensitive walkway and a pressure-sensitive treadmill for a group of healthy dogs. Although there was some concern that dogs (or people) might have different gait patterns over ground, compared with the patterns detected when walking or running on a treadmill, 1 study11 of 5 mixed-breed dogs found no significant differences between overground gait patterns and those assessed on a treadmill. An investigation of 40 adult human subjects also indicated that differences in most gait variables are no longer detectable after approximately 6 to 7 minutes of acclimatization to a treadmill.6
The purpose of the study reported here was to assess the repeatability of gait analysis results for healthy dogs determined by use a of pressure-sensitive treadmill designed to measure temporospatial and kinetic gait variables in this species. We hypothesized that the instrumented treadmill would provide repeatable intrasession and intersession measurements of kinetic and temporospatial variables in the study sample.
Materials and Methods
Dogs
Healthy dogs were prospectively enrolled in the study. Recruitment for study participation was performed by 1 investigator (KAH). The study dogs comprised a convenience sample of privately owned dogs recruited over a 2-week period from this investigator's clinical practice. Dogs were recruited by word of mouth, and all of the clients were verbally informed about the study goals and procedures by investigators (KAH or DB) prior to enrollment. To be included in the study, dogs were required to be skeletally mature, with no evidence of clinical lameness on the basis of physical examination results and owner feedback. Previous exposure to a treadmill was not a criterion for inclusion in the study. Dogs with evidence of ataxia or that were not considered fit enough to walk or trot on a treadmill were excluded. All dog owners provided informed consent for study participation. The study protocol was reviewed and approved by the University of Cambridge Ethics and Welfare Committee.
Pressure-sensitive treadmill
All measurements were conducted with 1 instrumented treadmill.a The overall dimensions of the treadmill unit were 210 × 65 × 80 cm (length × width × height), and the walking surface was 190 × 46 cm (length × width). The sensors were calibrated before each run, in accordance with the manufacturer's instructions. The pressure-sensitive surface was 163 × 41 cm (length × width) and included 9,216 high-precision, individually calibrated capacitive pressure sensors (each with a spatial resolution of 0.4 cm) with a collection frequency of 100 Hz. The treadmill was operable at speeds from 0 to 12 km/h, adjustable in 0.1-km/h increments, and was designed for dogs with body weights of 5 to 80 kg.22
Investigation of potential treadmill speed effects on GRF measurements
Although static measurement of pressures and collection of force data are relatively uncomplicated, introduction of a moving belt above the pressure-sensing surface adds a layer of complexity. To investigate whether this influenced pressure distribution (GRF) measurements, a test device was created in which a pair of bicycle wheels was connected through an axle with a platform suspended below the axle for the addition of weights. The device was lowered onto the treadmill belt so that the tires were in contact with the part of the treadmill belt over the pressure-sensing surface and bearing the full device weight, and the wheels were allowed to rotate freely while the device was kept in place with supports mounted on either side of the axle (Figure 1). The pressure profiles created by contact between the tires and the belt were captured and analyzed with the treadmill at a standstill (0 km/h) and as the speed of the treadmill was progressively increased from 0 to 1.3 to 1.5 km/h, then in increments of 0.5 km/h to a maximum of 7.5 km/h. Three trials/weight/speed setting were recorded.
Schematic drawing depicting a custom test device developed for assessment of the potential influence of treadmill speed on GRF measurements. The pressure-sensitive treadmilla was designed for gait analysis of dogs with body weights of 10 to 80 kg.22 A pair of bicycle wheels was connected by a central axle, and a platform was added for the attachment of weights. The wheels were lowered onto the surface of the treadmill and kept in place by the supports as pressure created at the resultant contact areas was measured in real time. Recordings were made across a range of predetermined speeds, and the force applied through the wheels was varied by suspending calibrated 2- or 3-kg weights below the midpoint of the axle. The total amount of weight applied ranged from 9.05 kg (for the device without added weights) to 16.05 kg.
Citation: American Journal of Veterinary Research 81, 12; 10.2460/ajvr.81.12.922
Data captured from individual pressure sensors were used to compute the mean pressure, peak pressure, and area of pressure contact, and the force applied across the pressure-sensing surface was calculated automatically. After collecting measurements for the 9.05-kg device at the described speeds without weights added, calibrated 2- and 3-kg weights were placed on the platform between the wheels, and the measurements were repeated at the same speeds with a range of added weights from 2 to 7 kg (for a range of total device weights from 9.05 to 16.05 kg). The final pressure distribution measurements, reported as GRFs, were normalized to the total weight of the device.
One client-owned dog was then acclimated to the treadmill and used for measurement of pressures (and calculation of GRFs, which were then normalized to the dog's body weight) when the treadmill was stationary and across a range of clinically relevant speeds (3 to 7 km/h [0.83 to 1.94 m/s], at 0.5 km/h increments). Three 10-second data captures were recorded at each speed, and the mean values for all 4 limbs were used for subsequent analysis. The acclimation and handling procedures for this dog were the same as described for the gait analysis protocol. This dog only participated in this phase of the study.
The weight-normalized force measurements and changes with increasing treadmill speed were reported for these experiments. Statistical analysis for associations between treadmill speed and measured variables was not performed.
Gait analysis protocol
For the gait analysis sessions, data were collected with dogs walking at a constant, predetermined speed of 3 km/h.14 This standardized speed was used for all dogs, regardless of size, body weight, or breed. Temporospatial (step length, stride length, duration of swing and stance phases, and cadence) and kinetic (GRF) data were recorded for all 4 limbs.
The treadmill was located centrally in the gait analysis room of the facility where the study took place and remained in the same position throughout the study. The same individuals (KAH and DB) handled the dogs for all examinations. Harnesses or leads were used for all dogs on the treadmill, and food treats were used as needed to encourage ambulation. Once the dogs were familiarized with the treadmill, they were allowed 2 minutes to acclimate to walking on the belt at a fixed speed of 3 km/h. After an additional 2 minutes of steady-state walking at the fixed speed, recording started. Three recordings, each of 10 seconds, were made during each trial. Trials in which there was an obvious change in gait pattern during the 10-second recording were excluded, and a replacement trial was performed.
Intrasession repeatability of the measurements was determined by comparing measurements for 2 trials ≤ 2 hours apart on the same day. Intersession repeatability was determined by comparing results for the first trial on the first study day (session 1) with results for a single trial performed between 4 and 7 days later (session 2).
Within each of the 3 treadmill trials, data from the three 10-second recordings were collated, and the mean of those measurements was used for data analysis. For each dog, GRF values were normalized to body weight.
Statistical analysis
Normality of the data was assessed with the Shapiro-Wilk test. Results are presented as mean and SD for normally distributed data or median and IQR for nonnormally distributed data.
Intrasession and intersession repeatability of temporospatial and kinetic measurements were characterized by calculation of the ICC. Two-way models were used, and the repeatability between each pair of measurements was estimated. The test of consistency was chosen, and intrasession or intersession repeatability of the measurements was judged on the basis of the ICC value with criteria as follows23: excellent (> 0.9), good (0.75 to 0.9), moderate (0.5 to 0.74), or poor (< 0.5). Bland-Altman analysis was used to identify potential outliers and determine the limits of agreement for intrasession or intersession measurements of all temporospatial gait variables. The difference ratio, expressed as a percentage, was also determined as a measure of repeatability (calculated as [difference of the 2 measurements/mean of the 2 measurements] × 100).
All analyses were conducted with an open-source software program.b Values of P < 0.05 were considered significant for all tests.
Results
Dogs
The main study sample included 15 dogs (8 males and 7 females). The median age was 85 months (range, 23 to 159 months; IQR, 29 to 114.5 months) and median weight was 22 kg (range, 3.7 to 31.2 kg; IQR, 16.5 to 29.5 kg). Most of the animals were mixed-breed dogs (n = 8). The remainder included 3 Whippets, 1 German Shepherd Dog, 1 Golden Retriever, 1 Labrador Retriever, and 1 Staffordshire Bull Terrier. Seven dogs had been on noninstrumented treadmills before the study, and 8 had not been on any type of treadmill before the study. The dog that was only included in the investigation of potential treadmill speed effects on GRF measurements was a 5-year-old 26-kg spayed female Labrador Retriever.
Changes in force measurements with increasing treadmill speed
Visual comparison of the weight-normalized force measurements for the custom test device revealed that the pressure sensed by the treadmill decreased as speed of the treadmill increased (Figure 2). Across the range of treadmill speeds tested (1.3 to 7.5 km/h), there was a mean 9.1% reduction in the force data (range, 2.6% to 16.6%), compared with the result when the treadmill was stationary (assigned a value of 1.0). The decreases in measurements were observed at all weights tested, with greater error occurring at greater speed, and the weight-normalized results were indicative of a systematic error in the measurements.
Mean ± SD weight-normalized forces recorded with the custom-made bicycle wheel testing device depicted in Figure 1 on an instrumented treadmill at a standstill (0 km/h) and at increasing treadmill speeds from 1.3 to 7.5 km/h (white circles) and body weight–normalized GRFs recorded for 1 healthy dog across the range of clinically relevant treadmill speeds (3 to 7 km/h; black squares). For the test device, the force values were calculated for 3 recordings/weight at each treadmill speed. For the dog, the GRF values were calculated for all 4 limbs in three 10-second recordings at each treadmill speed. For the testing device, there was a mean 9.1% reduction (range, 2.6% to 16.6%) in the magnitude of force data over the full range of speeds tested. For the dog, there was a mean 12.4% (range, 8.8% to 16.0%) decrease in GRF over the range of speeds tested. With normalization for weight, differences from a value of 1.0 represent errors in the measurements.
Citation: American Journal of Veterinary Research 81, 12; 10.2460/ajvr.81.12.922
In the session where data were collected for 1 healthy dog at treadmill speeds of 3 to 7 km/h, results were similar; body weight–normalized GRFs decreased by a mean of 12.4% (range, 8.8% to 16.0%), compared with the value obtained when the treadmill belt was stationary (1.0).
Repeatability analysis
Analysis results for intrasession and intersession repeatability (ICCs for the consistency test) were summarized (Tables 1 and 2, respectively). Overall, temporospatial measurements had higher ICCs (greater repeatability) than kinetic (GRF) measurements. For all temporospatial measurements, intrasession repeatability was excellent (ICCs > 0.9). Intersession repeatability for stride length, cadence, and step length was excellent, and intersession repeatability for the remaining temporospatial variables was good (ICCs ≥ 0.75 and < 0.9). Ground reaction forces for the pelvic limbs had excellent intrasession repeatability (ICCs > 0.9), with moderate (ICC of 0.72 for the right side) or good (ICC of 0.78 for the left side) intersession repeatability, whereas those for the thoracic limbs had moderate intrasession repeatability (ICCs of 0.68 and 0.72 for the left and right sides, respectively) and poor intersession repeatability (ICCs of 0.35 and 0.36 for the left and right sides, respectively).
Results of ICC and difference ratio analysis for assessment of intrasession repeatability of temporospatial and kinetic variable measurements obtained from 15 healthy skeletally mature dogs during 2 gait analysis trials on a pressure-sensitive treadmill ≤ 2 hours apart on the same day.
| All dogs (n = 15) | Dogs > 15 kg (n = 12) | |||
|---|---|---|---|---|
| Variable | ICC (95% CI) | Difference ratio (median [IQR] %) | ICC (95% CI) | Difference ratio (median [IQR] %) |
| Stride length (cm) | 0.929 (0.761 to 0.977) | 1.876 (−1.28 to 5.032) | 0.951 (0.838 to 0.986) | 1.855 (−1.329 to 5.039) |
| Cadence (steps/min) | 0.904 (0.738 to 0.967) | 2.089 (−0.971 to 5.149) | 0.928 (0.771 to 0.979) | 1.9 (−1.107 to 4.907) |
| Step length (cm) | ||||
| Left thoracic limb | 0.911 (0.763 to 0.969) | 2.317 (−0.84 to 5.482) | 0.899 (0.687 to 0.97) | 2.277 (−0.798 to 5.352) |
| Right thoracic limb | 0.913 (0.580 to 0.975) | 4.131 (−0.417 to 8.679) | 0.927 (0.767 to 0.978) | 3.968 (−0.89 to 8.225) |
| Left pelvic limb | 0.927 (0.799 to 0.975) | 3.448 (0.342 to 6.555) | 0.9 (0.69 to 0.97) | 3.366 (1.385 to 5.348) |
| Right pelvic limb | 0.923 (0.702 to 0.976) | 2.691 (−0.232 to 5.614) | 0.964 (0.881 to 0.99) | 3.388 (0.584 to 6.193) |
| Stance phase (%) | ||||
| Left thoracic limb | 0.961 (0.891 to 0.987) | 0.582 (0.062 to 1.103) | 0.791 (0.423 to 0.935) | 0.546 (−0.167 to 1.259) |
| Right thoracic limb | 0.971 (0.916 to 0.99) | 1.198 (0.831 to 1.566) | 0.916 (0.735 to 0.975) | 0.779 (0.459 to 1.099) |
| Left pelvic limb | 0.957 (0.869 to 0.986) | 0.987 (0.151 to 1.824) | 0.852 (0.565 to 0.955) | 0.951 (0.049 to 1.853) |
| Right pelvic limb | 0.955 (0.867 to 0.985) | 0.785 (0.001 to 1.565) | 0.965 (0.883 to 0.99) | 0.684 (0.157 to 1.212) |
| Swing phase (%) | ||||
| Left thoracic limb | 0.963 (0.897 to 0.987) | 0.840 (0.189 to 1.491) | 0.811 (0.467 to 0.942) | 0.804 (−0.314 to 1.922) |
| Right thoracic limb | 0.971 (0.916 to 0.990) | 1.690 (1.105 to 2.275) | 0.917 (0.738 to 0.975) | 1.511 (0.918 to 2.105) |
| Left pelvic limb | 0.957 (0.875 to 0.986) | 1.552 (0.242 to 2.863) | 0.864 (0.595 to 0.959) | 1.804 (0.701 to 2.907) |
| Right pelvic limb | 0.953 (0.857 to 0.984) | 1.447 (0.059 to 2.836) | 0.963 (0.878 to 0.989) | 1.388 (0.089 to 2.688) |
| GRF (N/kg) | ||||
| Left thoracic limb | 0.676 (0.288 to 0.876) | 1.535 (0.158 to 2.913) | 0.862 (0.59 to 0.958) | 1.937 (0.487 to 3.387) |
| Right thoracic limb | 0.716 (0.351 to 0.894) | 2.715 (1.4 to 4.03) | 0.899 (0.569 to 0.97) | 2.845 (1.628 to 4.062) |
| Left pelvic limb | 0.967 (0.905 to 0.989) | 3.306 (2.03 to 4.582) | 0.878 (0.631 to 0.963) | 3.325 (2.246 to 4.405) |
| Right pelvic limb | 0.954 (0.872 to 0.984) | 3.618 (0.804 to 6.432) | 0.863 (0.593 to 0.958) | 3.671 (1.359 to 5.983) |
Each trial comprised three 10-second recordings for each dog, and the mean of the 3 measurements/variable was used for analysis. The test of consistency was chosen for ICC analysis. The difference ratio, expressed as a percentage, was calculated as follows: (difference of the 2 measurements/mean of the 2 measurements) × 100.
Results of ICC and difference ratio analysis for assessment of intersession repeatability of temporospatial and kinetic variable measurements obtained from the dogs in Table 1 during 2 gait analysis sessions on the same pressure-sensitive treadmill 4 to 7 days apart.
| All dogs (n = 15) | Dogs > 15 kg (n = 12) | |||
|---|---|---|---|---|
| Variable | ICC (95% CI) | Difference ratio (median [IQR] %) | ICC (95% CI) | Difference ratio (median [IQR] %) |
| Stride length (cm) | 0.961 (0.890 to 0.986) | 4.599 (2.006 to 7.192) | 0.863 (0.592 to 0.958) | 3.356 (0.892 to 5.82) |
| Cadence (steps/min) | 0.984 (0.953 to 0.995) | 3.161 (1.22 to 5.102) | 0.855 (0.572 to 0.956) | 3.418 (1.072 to 5.765) |
| Step length (cm) | ||||
| Left thoracic limb | 0.952 (0.864 to 0.983) | 5.962 (3.254 to 8.671) | 0.827 (0.506 to 0.947) | 5.714 (2.841 to 8.588) |
| Right thoracic limb | 0.939 (0.830 to 0.979) | 3.218 (−0.044 to 6.48) | 0.792 (0.425 to 0.935) | 3.525 (0.541 to 6.51) |
| Left pelvic limb | 0.942 (0.839 to 0.980) | 2.038 (−1.451 to 5.527) | 0.836 (0.526 to 0.95) | 2.348 (−0.832 to 5.528) |
| Right pelvic limb | 0.946 (0.839 to 0.982) | 2.930 (0.158 to 5.703) | 0.848 (0.556 to 0.954) | 2.587 (−0.228 to 5.402) |
| Stance phase (%) | ||||
| Left thoracic limb | 0.820 (0.553 to 0.935) | 0.977 (0.182 to 1.772) | 0.785 (0.41 to 0.933) | 0.936 (0.473 to 1.400) |
| Right thoracic limb | 0.775 (0.465 to 0.917) | 0.661 (−0.126 to 1.448) | 0.867 (0.602 to 0.96) | 0.651 (0.227 to 1.075) |
| Left pelvic limb | 0.804 (0.494 to 0.930) | 2.036 (1.255 to 2.818) | 0.903 (0.699 to 0.971) | 1.483 (0.751 to 2.216) |
| Right pelvic limb | 0.808 (0.514 to 0.932) | 1.621 (0.373 to 2.869) | 0.957 (0.859 to 0.988) | 0.831 (−0.006 to 1.668) |
| Swing phase (%) | ||||
| Left thoracic limb | 0.822 (0.559 to 0.936) | 1.622 (−0.004 to 3.248) | 0.79 (0.42 to 0.935) | 1.584 (0.706 to 2.463) |
| Right thoracic limb | 0.777 (0.469 to 0.918) | 1.338 (0.11 to 2.567) | 0.876 (0.627 to 0.963) | 1.297 (0.751 to 1.844) |
| Left pelvic limb | 0.806 (0.509 to 0.931) | 2.762 (1.509 to 4.016) | 0.919 (0.744 to 0.976) | 2.119 (0.998 to 3.241) |
| Right pelvic limb | 0.808 (0.514 to 0.932) | 2.396 (0.509 to 4.284) | 0.958 (0.86 to 0.988) | 1.499 (0.18 to 2.818) |
| GRF (N/kg) | ||||
| Left thoracic limb | 0.351 (−0.140 to 0.716) | 2.545 (−0.402 to 5.492) | 0.863 (0.594 to 0.959) | 2.222 (0.784 to 3.661) |
| Right thoracic limb | 0.355 (−0.143 to 0.719) | 3.491 (−0.456 to 7.438) | 0.832 (0.517 to 0.949) | 2.322 (0.888 to 3.756) |
| Left pelvic limb | 0.779 (0.46 to 0.92) | 8.403 (5.138 to 11.669) | 0.835 (0.524 to 0.95) | 6.330 (4.106 to 8.555) |
| Right pelvic limb | 0.72 (0.339 to 0.897) | 7.004 (0.288 to 13.72) | 0.7 (0.241 to 0.903) | 4.884 (1.547 to 8.222) |
For intersession analysis, data from the first of 2 trials on the first study day (session 1) were compared with data obtained on the subsequent date (session 2).
See Table 1 for remainder of key.
On visual examination of the data, dogs that weighed < 15 kg appeared to have greater variability in measurements of thoracic limb GRFs between sessions. Three dogs with body weights < 15 kg had measurement differences that were close to or exceeded the limits of agreement for intersession measurements on Bland-Altman plots and were deemed outliers; the plot for GRF data is shown (Figure 3). In contrast, dogs that weighed > 15 kg had similar measurement differences regardless of the mean values. These 3 dogs were not automatically excluded from the analysis, but a second analysis was run in which they were excluded. Removal of the 3 dogs that weighed < 15 kg from the analysis resulted in greater intrasession and intersession ICCs for GRFs of both thoracic limbs and greater intersession ICC for GRF of the left pelvic limb; all GRF values for the data set of dogs > 15 kg (n = 12) had good repeatability (ICCs > 0.75 and < 0.9), except in the intersession analysis for the right pelvic limb, which had moderate repeatability (ICC, 0.7; Tables 1 and 2).
Bland-Altman plots depicting intersession agreement of mean weight-normalized GRF measurements for the left (A) and right (B) thoracic limbs of 15 healthy skeletally mature dogs. Intersession comparisons were made with data collected from the first of 2 trials on the first study day (session 1) and from a single trial performed 4 to 7 days later (session 2). Each trial consisted of three 10-second recordings, and the mean of the 3 measurements was used for analysis. Each data point represents results for 1 dog. The x-axis represents the mean of measurements for the 2 sessions, and the y-axis represents the difference between measurements for the 2 sessions. The solid horizontal line depicts the mean intersession difference (bias) in measurements, and shading indicates the 95% CI of the mean difference. The dotted lines indicate the upper and lower limits of agreement (1.96 SDs of the mean difference). Measurements for 3 dogs that weighed < 15 kg (each indicated by a unique number with body weight shown next to the data point) and were considered outliers are indicated.
Citation: American Journal of Veterinary Research 81, 12; 10.2460/ajvr.81.12.922
Difference ratio calculations yielded median values < 5% for most temporospatial and kinetic measurements. Values > 5% were found on intersession comparison of step length for the left thoracic limb (5.96%) and intersession comparison of GRF for the left and right pelvic limbs (8.4% and 7%, respectively, for the data set of all dogs).
Discussion
The present study revealed excellent intrasession repeatability and good to excellent intrasession repeatability of temporospatial gait variable measurements for healthy dogs in this sample, whereas kinetic (GRF) measurements were more variable, particularly for intersession comparisons. As such, the data partially supported the study hypothesis. Although there was a systematic effect of treadmill speed on weight-normalized force measurements, the results suggested that in a clinical context, the error can be ignored as long as the data are captured at the same speed in each recording session. This is an advantage of the treadmill system, compared with a standard pressure-sensitive walkway; although an experienced handler can minimize variability in the velocity at which a dog moves over the walkway, this variable cannot be controlled as accurately as the belt speed of an appropriately calibrated treadmill. It is common practice to set a range of velocities that will be acceptable for gait analysis with walkway-type systems, but published values for these recommended ranges can vary by up to 50%.24
With regard to the technical performance of the treadmill system, the results of the present study were encouraging. There does not appear to be any consensus on the degree of repeatability needed for a gait analysis system to be clinically useful, but the authors of a recent report25 suggested that gait analysis systems for use in people should have the ability to detect changes of < 5%. In the present study, the difference ratio (essentially the percentage error between 2 recordings for the same dog) was < 5% for nearly all of the gait variables, supporting the repeatability of most objective gait analysis measures with this system. There were only 3 variables for which this value was > 5% in the overall study sample: intersession comparisons of step length of the left thoracic limb (5.96%), GRF of the left pelvic limb (8.4%), and GRF of the right pelvic limb (7%). Although, to the authors’ knowledge, there are no published guidelines for acceptable difference ratios, authors of a previous investigation with a human treadmill system reported values in the range of 1% to 3%.26
Although the use of a pressure-sensitive treadmill for gait assessment of dogs has been previously described, the report by Assaf et al21 described a comparison between a pressure-sensitive treadmill system and a pressure-sensitive walkway. The goal of our study was to determine the repeatability of measurements made with the instrumented treadmill under clinical conditions. The technical performance of several systems designed for human use has been evaluated with similar test-retest protocols,18,27,28 and the results for the system designed for dogs and evaluated in the present study were broadly similar.
Three major concerns are typically cited for the use of pressure-sensitive treadmill systems for canine gait assessment: cost, noncompliance of the dogs, and lack of relevance to overground walking. The system that was evaluated in the present study was less expensive than the pressure-sensitive walkway and slightly more expensive than the single force plate–based system in our gait laboratory. In our hands, data capture and analysis were completed in approximately 3 to 5 minutes with the instrumented treadmill, which compared favorably with the 22 to 46 minutes required for pressure-sensitive walkway analysis in the personal experience of 2 authors (KAH and MJA). Most of the time difference was identified in the data analysis, with the instrumented treadmill having software that allowed for automated footfall identification and data output in a matter of seconds.
The issue of compliance is important, but this was not an obstacle in the present study. Half of the dogs in this study had not previously been on a treadmill, but all of the dogs were able to walk at a steady state on the treadmill at 3 km/h for the required 10-second intervals after only a short period of acclimatization. For clinical trials, there may be some benefit in undertaking more extended acclimatization prior to study enrollment (a study of human subjects suggests 7 minutes might be a useful guideline), but this was not evaluated in our study.6
The literature in human and veterinary medicine is inconsistent on the issue of alterations in gait patterns when treadmill-based gait analysis is undertaken. It is generally accepted that gait patterns are different when people or dogs walk on a treadmill, compared with those observed over level ground,11,29 but there is no clear consensus as to whether these differences have a negative impact on the clinical relevance of treadmill data as indicators of changes in gait over time following surgical or medical interventions. Recent literature in this regard appears to show that people perform so-called cautious walking on a treadmill.30,31 Temporal gait variables such as stride length and cadence did not vary significantly between treadmill and overground analyses in healthy people in those studies, but joint movement and muscle activation patterns showed more variety. Overall, walking on a treadmill was considered an appropriate means to evaluate gait under the same circumstances each time.30,31
It is also important to recognize the advantages offered by the instrumented treadmill used in the present study. It is reasonable to assume that the collection of a larger set of data provides for a more robust representation of the gait for an individual dog. For a medium-sized dog at a walk, data from up to 40 complete gait cycles could be collected over the course of the (total) 30-second recording period (not shown). This compared favorably with the 6 to 10 complete cycles that typically result from 6 acceptable passes over a pressure-sensitive walkway, although clinical research would be needed to determine whether this is associated with greater sensitivity for detection of gait abnormalities. Another advantage of the instrumented treadmill over the force plate and pressure-sensitive walkways used in the authors’ laboratory related to the use of automated processes to identify and correctly label footfalls. The system included a live-streaming video feed that could be used to confirm the identity of individual feet as they landed on the treadmill, and the fact that the dog's position over the treadmill remained fairly static over time meant that the software discriminated left from right and thoracic from pelvic limbs without any human input. The algorithms in the software objectively determined heel strike, toe off, total sensor area, and other relevant features.
The primary limitation of the study reported here was the use of a small and heterogeneous population of dogs. We chose 2 approaches to analyze the data; data from dogs that weigh > 15 kg are representative of the population of dogs most often presented to the authors for gait analysis, and the use of these dogs allowed us to evaluate the instrument within its recommended weight range. However, we were also interested in determining whether smaller dogs could be encouraged to walk on the treadmill and, if so, whether the treadmill would detect their footfalls and produce meaningful temporospatial and kinetic data. Even though the system is factory-optimized for dogs that weigh 10 to 80 kg,22 it was possible to record data from smaller dogs. The use of 1 treadmill speed, irrespective of body shape or size, was reasonable for most of the breeds included but this may have had a negative impact on the quality of the data for dogs < 15 kg. Further research is needed to assess the potential impact of body shape and size on the variables assessed in the present study and determine whether optimal treadmill speeds can be identified for dogs of various breeds and sizes. Another important limitation of the present study was that it was conducted with healthy dogs that had no clinical evidence of lameness. The question of whether the instrumented treadmill is sufficiently accurate for the detection of lameness in dogs and assessment of changes in lameness over time was not explored, and further investigations are needed to answer this question.
Acknowledgments
The instrumented treadmill was provided by Zebris Medical GmbH. No direct financial support was provided for this study. Dr. Häusler serves as a consultant for Zebris Medical GmbH.
Presented in abstract form at the 10th International Symposium on Veterinary Rehabilitation and Physical Therapy, Knoxville, Tenn, 2018.
ABBREVIATIONS
| GRF | Ground reaction force |
| ICC | Intraclass correlation coefficient |
| IQR | Interquartile (25th to 75th percentile) range |
Footnotes
CanidGait, Zebris Medical GmbH, Isny im Allgäu, Germany.
R, version 3.3.0 for Mac, R Foundation for Statistical Computing, R Core Team, Vienna, Austria. Available at: www.R-project.org/. Accessed Apr 1, 2020.
References
1. McLaughlin RM. Kinetic and kinematic gait analysis in dogs. Vet Clin North Am Small Anim Pract 2001;31:193–201.
2. Waxman AS, Robinson DA, Evans RB, et al. Relationship between objective and subjective assessment of limb function in normal dogs with an experimentally induced lameness. Vet Surg 2008;37:241–246.
3. Quinn MM, Keuler NS, Lu Y, et al. Evaluation of agreement between numerical rating scales, visual analogue scoring scales, and force plate gait analysis in dogs. Vet Surg 2007;36:360–367.
4. Bockstahler BA, Skalicky M, Peham C, et al. Reliability of ground reaction forces measured on a treadmill system in healthy dogs. Vet J 2007;173:373–378.
5. Budsberg SC, Rytz U, Johnston SA. Effects of acceleration on ground reaction forces collected in healthy dogs at a trot. Vet Comp Orthop Traumatol 1998;11:15–19.
6. Meyer C, Killeen T, Easthope CS, et al. Familiarization with treadmill walking: how much is enough? Sci Rep 2019;9:5232.
7. Hobbs SJ, Clayton HM. Sagittal plane ground reaction forces, centre of pressure and centre of mass in trotting horses. Vet J 2013;198:e14–e19.
8. Lloyd R, Cooke C. Biomechanical differences associated with two different load carriage systems and their relationship to economy. Human Mov 2011;12:65–74.
9. von Porat A, Henriksson M, Holmström E, et al. Knee kinematics and kinetics in former soccer players with a 16-year-old ACL injury—the effects of twelve weeks of knee-specific training. BMC Musculoskelet Disord 2007;8:35.
10. Stejskal M, Torres BT, Sandberg GS, et al. Variability of vertical ground reaction forces collected with one and two force plates in healthy dogs. Vet Comp Orthop Traumatol 2015;28:318–322.
11. Torres BT, Moëns NMM, Al-Nadaf S, et al. Comparison of overground and treadmill-based gaits of dogs. Am J Vet Res 2013;74:535–541.
12. Zeni JA Jr, Richards JG, Higginson JS. Two simple methods for determining gait events during treadmill and overground walking using kinematic data. Gait Posture 2008;27:710–714.
13. Evans R, Gordon W, Conzemius M. Effect of velocity on ground reaction forces in dogs with lameness attributable to tearing of the cranial cruciate ligament. Am J Vet Res 2003;64:1479–1481.
14. Horstmann GA, Huethe F, Dietz V. Special treadmill for the investigation of standing and walking in research and in clinical medicine [in German]. Biomed Tech (Berl) 1987;32:250–254.
15. Kram R, Powell AJ. A treadmill-mounted force platform. J Appl Physiol 1989;67:1692–1698.
16. Firminger CR, Vernillo G, Savoldelli A, et al. Joint kinematics and ground reaction forces in overground versus treadmill graded running. Gait Posture 2018;63:109–113.
17. Giakas G, Baltzopoulos V. Time and frequency domain analysis of ground reaction forces during walking: an investigation of variability and symmetry. Gait Posture 1997;5:189–197.
18. Reed LF, Urry SR, Wearing SC. Reliability of spatiotemporal and kinetic gait parameters determined by a new instrumented treadmill system. BMC Musculoskelet Disord 2013;14:249.
19. Wearing SC, Reed LF, Urry SR. Agreement between temporal and spatial gait parameters from an instrumented walkway and treadmill system at matched walking speed. Gait Posture 2013;38:380–384.
20. Brebner NS, Moëns NMM, Runciman JR. Evaluation of a treadmill with integrated force plates for kinetic gait analysis of sound and lame dogs at a trot. Vet Comp Orthop Traumatol 2006;19:205–212.
21. Assaf ND, Rahal SC, Mesquita LR, et al. Evaluation of parameters obtained from two systems of gait analysis. Aust Vet J 2019;97:414–417.
22. CanidGait. Stance and gait analysis for dogs. Measuring system for diagnostics, therapy and rehabilitation. Available at: www.zebris.de/en/medical/products-solutions/canidgait-for-dogs/. Accessed Jul 2, 2020.
23. Koo TK, Li MY. A guideline of selecting and reporting intraclass correlation coefficients for reliability research (Erratum published in J Chiropr Med 2017;16:346). J Chiropr Med 2016;15:155–163.
24. Piazza AM, Binversie EE, Baker LA, et al. Variance associated with walking velocity during force platform gait analysis of a heterogeneous sample of clinically normal dogs. Am J Vet Res 2017;78:500–507.
25. McSweeney SC, Reed LF, Wearing SC. Reliability and minimum detectable change of measures of gait in children during walking and running on an instrumented treadmill. Gait Posture 2020;75:105–108.
26. Al-Amri M, Al Balushi H, Mashabi A. Intra-rater repeatability of gait parameters in healthy adults during self-paced treadmill-based virtual reality walking. Comput Methods Biomech Biomed Engin 2017;20:1669–1677.
27. Girard O, Brocherie F, Morin J, et al. Intrasession and intersession reliability of running mechanics during treadmill sprints. Int J Sports Physiol Perform 2016;11:432–439.
28. Donath L, Faude O, Lichtenstein E, et al. Validity and reliability of a portable gait analysis system for measuring spatiotemporal gait characteristics: comparison to an instrumented treadmill. J Neuroeng Rehabil 2016;13:6.
29. Riley PO, Dicharry J, Franz J, et al. A kinematics and kinetic comparison of overground and treadmill running. Med Sci Sports Exerc 2008;40:1093–1100.
30. Lee SJ, Hidler J. Biomechanics of overground vs. treadmill walking in healthy individuals. J Appl Physiol 2008;104:747–755.
31. Shi L, Duan F, Yang Y, et al. The effect of treadmill walking on gait and upper trunk through linear and nonlinear analysis methods. Sensors (Basel) 2019;19:2204.
