Materials and Methods

Animals—Five adult mixed-breed dogs (weight range, 20 to 30 kg) from an established research colony were evaluated in this study. All dogs had no physical or radiographically detectable pathological changes in the hip or stifle joints. For each dog, results of an initial force plate analysis, CBC, serum biochemical analysis, and complete physical examination performed prior to initiation of the study indicated no abnormalities. The dogs were housed indoors in a climate-controlled environment and fed commercially available dog food ad libitum. Use of these animals was approved by the University of Georgia Institutional Animal Care and Use Committee.

The number of dogs in the study group was determined on the basis of a power analysis to detect a 5% difference with an α error of 0.05 and a β error of 0.8 for the first 3 Fourier coefficients for the hip joint, the first 5 coefficients for the femorotibial joint, and the first 6 coefficients for the tarsal joint.¹¹ On the basis of a previous study¹¹ that used Fourier analysis, population size estimates for the hip and femorotibial joints were 5, whereas 16 animals were estimated to be needed for the tarsal joint.

Motion data collection—Thirty spherical retroreflective markers (diameter, approx 8 mm) were affixed with double-sided tape and cyanoacrylate to the right and left hind limbs and right and left sides of the pelvis (Appendix). A bilateral rigid-body segmental model of the canine hind limb and pelvis was used to collect kinematic data as described elsewhere.¹²

A 3-D testing space was established on a 13-m walkway. Right-handed orthogonal coordinate axes were used to describe the testing space in 3-D, with 0,0,0 (X,Y,Z) located in the center of the testing space. Prior to data collection on each testing day, the system was calibrated with a calibration frame^a of known dimensions and by dynamic linearization with a custommade 0.700-m wand. Marker locations were captured by a kinematic system of 8 infrared cameras^b arranged around the gait platform. Cameras captured sample data at 200 Hz. Data were recorded and analyzed by a motion analysis program.^c

Initially, a static data collection was performed for each dog. Four markers on both the right and left hind limbs were removed during subsequent dynamic data collections (Appendix). These markers were mathematically reconstructed from the initial static data and were used as virtual markers during the dynamic data collections.^12–17 This use of virtual markers was necessitated by limitations in marker visibility during walking or trotting as a result of the partial or complete truncal concealment of certain markers. All data for individual dogs were obtained during 1 testing period on 1 day.

Overground gait data were recorded as each dog moved through the calibrated space at a walk and trot. The order in which each gait was performed was identical for all dogs. Each dog was walked across the testing space at a speed of 0.9 to 1.2 m/s and trotted across the testing space at a speed of 1.7 to 2.1 m/s. Each gait was recorded 5 times for analysis. Passes in which the dog visibly changed velocity, turned its head, broke stride, or made any aberrant motions were discarded immediately.

Treadmill gait data were recorded with dogs moving on the treadmill at a walk and trot. All dogs underwent treadmill training every other day for approximately 2 weeks prior to study initiation. The order in which each gait was performed was identical for all dogs. Individual dogs were introduced gently onto the treadmill. Each dog was restrained with a standard harness that was loosely attached to the treadmill with a leash. The treadmill motion was initiated, and the speed was increased until a steady walk was achieved. A recording of the dog walking at a treadmill belt speed of 1.0 m/s was obtained. After approximately 10 seconds of steady ambulation at the defined speed, walking gait data were recorded over an interval of 20 seconds. The treadmill speed was then slowly increased until a steady trot was achieved. A recording of the dog trotting at a treadmill belt speed of 1.9 m/s was obtained. After approximately 10 seconds of steady ambulation at the defined speed, trotting gait data were recorded over an interval of 20 seconds. The first 5 complete gait cycles were used for analysis. The harness used for securing the dogs to the treadmill was in place on all dogs during each period of overground or treadmill testing.

Although data were collected for both sides of the body, data from 1 body side were used for comparisons. This was necessitated by considerable marker concealment and data loss for the side of the dog on which the handler was located during overground testing. This problem was not present during treadmill testing. Comparisons of data collected during overground and treadmill testing were performed with data obtained from the same limb (right or left) for each individual dog. All data (overground and treadmill) for each dog were collected on the same day.

Data analysis—Waveforms were generated during each gait cycle for the overground and treadmill testing. The waveforms were compiled graphically with 95% confidence intervals. These waveforms were then compared via GIFA^15,18 and Fourier analysis.^11,15,19 Significance was set at a value of P < 0.05.

Eight Fourier coefficients were used to characterize hip, stifle, and tarsal joint motions. Comparison of the overground and treadmill Fourier coefficients was accomplished with a paired t test. All hypothesis tests were 2-sided, and the significance level was α = 0.05. The paired t tests were performed with statistical analysis software.^d

Results

Sagittal (flexion and extension), transverse (internal and external rotation), and frontal (abduction and adduction) plane kinematics during movement of the distal segment relative to the proximal segment for each of the 3 joints (hip, femorotibial, and tarsal joints) were generated and collected from each dog during each dynamic gait cycle for overground and treadmill-based gaits at both a walk and a trot. Each plane of motion was evaluated independently for comparative analysis (Figure 1).

View Full Size

Graphs illustrating the mean (solid lines) joint angles with 95% confidence intervals (dashed lines) obtained for the hip (left column), femorotibial (central column), and tarsal (right column) joints of 5 dogs in a study of sagittal (flexion-extension [Flex-ext]), transverse (internal-external rotation [Int-ext]), and frontal (abduction-adduction [Abd-add]) plane kinematics during overground (black lines) or treadmill-based (red lines) ambulation. Quantitative angular change during movement of the distal segment relative to the proximal segment is indicated by the appropriate waveform. A—Values obtained during walking. B—Values obtained during trotting.

Citation: American Journal of Veterinary Research 74, 4; 10.2460/ajvr.74.4.535

• Download Figure
• Download figure as PowerPoint slide

Fourier analysis—No significant differences were found between overground and treadmill-based gaits during flexion and extension joint motion for each joint (hip, femorotibial, or tarsal joint) at both a walk and a trot (Table 1). Power calculations for the tarsal joint coefficients ranged from 0.23 to 0.89.

Significant (P < 0.05) differences were found for internal and external joint motion between overground and treadmill-based gaits as follows: the hip, femorotibial, and tarsal joints during trotting and the hip and tarsal joints during walking (Table 2). Significant (P < 0.05) differences were found for abduction and adduction joint motion between overground and treadmill-based gaits for the femorotibial joint during trotting and the hip and tarsal joints during walking (Table 3).

GIFA—Significant (P < 0.05) differences were found among all planes of motion (sagittal [flexion and extension], transverse [internal and external rotation], and frontal [abduction and adduction]) of overground and treadmill-based gaits for the hip, femorotibial, and tarsal joints during both walking and trotting.

Discussion

In the present study, data collection for both overground and treadmill-based gaits produced similar waveform shapes for the hip, femorotibial, and tarsal joints in dogs. However, comparison of these waveforms with 2 methods of waveform analysis provided varied results. Generalized indicator function analysis revealed significant (P < 0.05) differences between overground and treadmill-based gaits for all planes of motion and all joints. Fourier analysis revealed no significant differences between overground and treadmill-based gaits for the sagittal plane; however, differences were detected via Fourier analysis for the transverse and frontal planes of motion. The discrepancies between findings obtained via GIFA and Fourier analysis were attributable to fundamental differences in these analyses.¹⁵ Given the distinct similarities among waveform shapes for all joints in dogs of the study reported here, the clinical relevance of these differences is unclear.

To the authors' knowledge, this is the first study to compare complete hind limb kinematic data associated with overground and treadmill-based ambulation for both walking and trotting gaits in dogs. In human medicine, treadmill-based gait assessment is widely used. For research purposes, a distinct advantage is the ability to control variables such as lighting, surface, and velocity.^9,20,21 However, it has been shown that variability exists between overground and treadmill-based gaits of humans during walking²² and running.²³ In the present study, overground and treadmill-based gaits of dogs produced similar waveform shapes for the hip, femorotibial, and tarsal joints during both walking and trotting. However, the detected variability in the various planes of motion was dependent on the analysis method, indicating that this method is an important factor in assessing differences in kinematic gait waveforms.

Sagittal plane kinematics for the dogs' hip, femorotibial, and tarsal joints were unaffected by mode of ambulation when assessed by Fourier analysis in the present study. This may be explained by the quantitatively larger angular change in the sagittal plane of motion, compared with angular changes in the frontal and transverse planes. Interestingly, the femorotibial joint was the only joint for which no differences between overground and treadmill-based gaits in all planes of motion were found. However, this was only true when the dogs were walking. It is possible that lower treadmill belt speeds may more closely mimic overground ambulation by limiting the effect of belt motion on ambulation.

The effect of marker placement on gait assessment data has been studied. Although overall waveform shapes remain similar, a shift of the waveform in the vertical axis secondary to differences in marker placement can occur.^11,24 Analysis methods such as Fourier analysis can be affected by this translation.¹⁵ Therefore, previous studies^11,24 of Fourier analysis have used a normalization procedure to decrease the impact of this shift on subsequent analysis. In the present study, in which no marker loss or reapplication occurred, a normalization procedure was not performed and Fourier analysis revealed no differences in sagittal plane motion. Therefore, the differences detected by GIFA were attributable to variations in the waveform shapes produced during overground and treadmill-based gaits.

In the present study, a limitation was sample size with respect to Fourier analysis of the tarsal joint. The number of dogs used was not adequate to rule out the possibility of generating a type II error in the tarsal joint data analysis, particularly when evaluating all of the coefficients required to reconstruct 95% of the waveform.¹¹

Although differences in overground and treadmill-based gaits of humans have been described, it has been argued that if the treadmill belt speed is constant and similar to overground velocity, then there should be no biomechanical differences between the 2 modes of ambulation.²⁵ However, results of experimental studies^22,23,26,27 have indicated that differences exist. In 1998, Savelberg et al²⁷ concluded that intrastride belt-speed variation can lead to kinematic differences between overground and treadmill gaits. Additionally, these differences are related to the overall power of the treadmill and the mass of the subject. In the present study, overground gait data obtained at a predetermined narrow velocity range were accepted for evaluation; however, belt speed was a constant for treadmill-based testing. It is possible that the differences detected by GIFA and, to a lesser extent, Fourier analysis may be secondary to intrastride belt-speed variations or minor differences between a variable overground velocity and constant treadmill belt speed. Further study of such differences is warranted.

Treadmill-based gait has been shown to alter joint range of motion.^26,28 Lee and Hidler²⁶ evaluated human gaits during overground and treadmill-based walking, and evaluation of sagittal plane kinematics revealed a decreased range of motion for the knees during treadmill walking. This finding is supported by results of another study²⁸ in humans, which also indicated that there was decreased joint range of motion during treadmill-based gait. Interestingly, Lee and Hidler²⁶ found very few overall differences between walking overground or on treadmills and concluded that this was attributable to muscular adaptations (modifications in muscle activation and joint moments and powers) that occurred during the treadmill-based gait, which resulted in similar joint kinematics for overground and treadmill ambulation. Similarly, when data obtained from dogs were assessed via Fourier analysis in the present study, no differences were found between overground and treadmill-based gaits during walking for the femorotibial joint. However, this was not true for the hip and tarsal joints. Nevertheless, overall waveform shapes were similar for all 3 joints and all planes of motion during walking and trotting.

Habituation is an integral part of treadmill use for gait analysis. Humans and other animals require sufficient time and training to become accustomed to treadmill-based gaits.^21,22,29 Matsas et al²⁹ found that for humans, treadmill-based walking could be generalized to overground walking after 6 minutes of treadmill use, indicating that a period of familiarization may be needed to produce comparable gaits. In another study²¹ in dogs, a 3-week acclimation period was allowed prior to data collection. In the present study, the dogs were trained for approximately 10 minutes every other day over a period of 2 weeks prior to study initiation, and data acquisition was obtained after 10 seconds of symmetric gait at the predetermined belt speed for walking and trotting. The similarity of resultant waveform shapes for all 3 joints in this study suggested that habituation occurred.

The hypothesis for the present study was supported by the findings. Differences between overground and treadmill-based gaits of dogs were detected; however, the ability to detect differences varied with joint, gait, and analysis method. The results of this study indicated that comparable hind limb kinematic waveform shapes for the hip, femorotibial, and tarsal joints can be acquired from dogs that are walking or trotting overground or on a treadmill. Furthermore, 3-D femorotibial kinematic gait data collected during walking as well as complete hind limb sagittal plane kinematic gait data collected during walking and trotting were comparable. The findings of the present study also confirmed that habituation can occur in the previously reported time frame.²¹ However, differences in analysis methods may alter the ability to detect differences between modes of ambulation. Although differences were found between the methods of ambulation in the present study, the clinical relevance of these differences has yet to be elucidated.