Gait Analysis Methods for Rodent Models of Arthritic Disorders: Reviews and Recommendations

Abstract

Gait analysis is a useful tool to understand behavioral changes in preclinical arthritis models. While observational scoring and spatiotemporal gait parameters are the most widely performed gait analyses in rodents, commercially available systems can now provide quantitative assessments of spatiotemporal patterns. However, inconsistencies remain between testing platforms, and laboratories often select different gait pattern descriptors to report in the literature. Rodent gait can also be described through kinetic and kinematic analyses, but systems to analyze rodent kinetics and kinematics are typically custom made and often require sensitive, custom equipment. While the use of rodent gait analysis rapidly expands, it is important to remember that, while rodent gait analysis is a relatively modern behavioral assay, the study of quadrupedal gait is not new. Nearly all gait parameters are correlated, and a collection of gait parameters is needed to understand a compensatory gait pattern used by the animal. As such, a change in a single gait parameter is unlikely to tell the full biomechanical story; and to effectively use gait analysis, one must consider how multiple different parameters contribute to an altered gait pattern. The goal of this article is to review rodent gait analysis techniques and provide recommendations on how to use these technologies in rodent arthritis models, including discussions on the strengths and limitations of observational scoring, spatiotemporal, kinetic, and kinematic measures. Recognizing rodent gait analysis is an evolving tool, we also provide technical recommendations we hope will improve the utility of these analyses in the future.

Introduction

Technological advances have made gait analysis a widely available tool for rodent models; however, gait analysis, itself, is not a new practice. Aristotle wrote on human motion, and scientists such as Boerhaave, Euler, and Carlet advanced the understanding of gait mechanics throughout the 18^th and 19^th centuries. In the late 1800s, human and quadrupedal gait patterns were famously recorded by Muybridge, and Hildebrand plots have been standard descriptors of the temporal gait sequence of quadrupeds since the 1960s ^{1, 2}. In recent years, gait technologies have continued to improve, and with advances in high-speed videography and force plates, gait is now possible in smaller organisms, including preclinical arthritis models.

In the clinical assessment of arthritis, gait analysis can be effectively used to assess mobility and function, and while the quadrupedal gait patterns used by rodents clearly vary from bipedal human patterns, the conceptual basis for gait analysis remains analogous - an altered gait pattern can be used to protect an injured limb from loading and/or movement-evoked pain. As such, several compensatory patterns are shared between quadrupeds and bipeds, including shuffle-stepping and limping ³. The frequent challenge in rodent gait analysis is that these compensatory patterns can be difficult to detect due to the ability to re-distribute load to three limbs rather than one and the rapidity of the gait sequence. Thus, although rodent gait varies significantly from human gait, it remains important to advance technologies for rodent and quadrupedal gait analysis.

In addition, bench to bedside translation in arthritis historically begins by testing new therapies in rodents before translating to larger organisms and clinical studies. However, gait analysis has followed the opposite trend, with sophisticated motion tracking and force measurements first being developed for humans and then scaled down to large animals and rodents. This trend is driven by scale advantages, with gait compensations being easier to detect in humans and large animals relative to rodents. In addition, rodents are prey animals, and evolution has likely conditioned rodents to mask signs of disability and pain ^4–7, making rodent gait compensations relatively more difficult to detect. Only recently have sophisticated gait tracking systems been applied to rodent arthritis models ^8–12, and even with these approaches, gait parameters are typically limited to the spatiotemporal pattern. Thus, even though quadrupedal gait has been studied for decades, sophisticated rodent gait analyses have only recently been applied to preclinical arthritis models. Nonetheless, use of gait analysis to study preclinical arthritis models will likely continue to expand.

In rodent models, gait is classified as a behavioral analysis. In her book, What’s Wrong With My Mouse? Behavioral Phenotyping of Transgenic and Knockout Mice, Jacqueline N. Crawley, Ph.D. takes care to inform scientists of the complexity of rodent behavioral analyses, stating:

“As in any field of science, behavioral research has evolved proper experimental designs and controls that must be correctly applied for the data to be interpretable. Little things, such as how to handle the mouse to reduce stress, can greatly affect the results of a behavioral experiment. Like microinjecting an oocyte or operating a DNA sequencer, the tricks of the trade are best learned from experts. You don’t want to waste your time reinventing the wheel.”¹³

Professor Crawley’s recommendation rings true for quadrupedal gait analysis as well. While the use of gait to assess rodent arthritis models is a relatively modern concept, the study of quadrupedal gait is not. Unfortunately, most commercial rodent gait analysis systems inundate users with large datasets – sometimes 50+ variables – to describe the rodent spatiotemporal gait pattern. Moreover, most of these variables are not independent and can be difficult to interpret without an understanding of quadrupedal gait ³. The goal of this review is to describe the state of rodent gait analysis and highlight fundamental gait parameters needed to accurately evaluate preclinical arthritis models. Rodent gait parameters include observational scores or measurements of spatiotemporal, kinetic, and kinematic gait characteristics. We structured this review with these sections, describing research conducted in these areas and the strengths and limitations of current approaches.

Observational Scoring

Terminology

Observational scoring uses rank-order scales to grade the severity of rodent gait abnormalities (Table 2). Since these scales have been developed by individual labs, scoring systems are relatively inconsistent, with some defining 0 as normal while others define the maximum as normal. Thus, when comparing results across laboratories, it is important to first understand the scale.

Table 2.

Review of Observational Gait Scores Used in Rodent Knee Arthritis Models

Species	Arthritis Model	Gait Method	Parameters Measured	Results	Therapies Tested:	Citation
Rats	CFA	Score	0: Normal 1: Guarding after noxious compression 2: Visible limping 3: No use of the hindlimb 4: No movement at all	Scores highest at day 1 and decreased over time. Still limping at day 21.	Morphine, Dexamethasone	14
Mice	CFA Carrageenan	Score	0: Normal 1: Moderate impairment of stance, toes contracted 2: Severe impairment, foot elevated, toes together	Carrageenan: Highest scores 3 hrs after injection. CFA: highest scores 24 hrs after injection.	Diclofenac, Morphine	20
Rats	CFA	Score	1: Normal .5: Marked limping 0: Three legged gait	Lame gait seen after injection	Diclofenac, Entada phaseoloides	19
Rats	Zymosan	Score	0: Normal 1: Mild disability 2: Difficulty walking 3: Three legged gait	Increased gait score after injection	Naproxen, ATB-346	18
Mice	Collagenase	Score	4: Normal 3: Minimal impairment 2: Moderate impairment 1: Significant impairment 0: Won’t walk on treadmill	Decreased score after injection	BoNT/B	17
Rat	Collagen Type II Antigen	Score	0: Normal 1: Guarding after noxious compression 2: Visible limping 3: No use of the hindlimb 4: No movement at all	Increased guarding scores up to day 14. Scores returned to control levels on day 21.	Enbrel (anti- TNF)	15
Rat	PGPS	Ink Prints	0: Normal 1: Slight limp, mainly toe prints on injured foot 2: Limping, only toes 3: Dragging and carrying leg, drag marks 4: Carrying leg entire time, no staining from injured foot	High scores (3.5) after induction of arthritis.	FX006 with TCA IR	16

CFA: Complete Freund’s Adjuvant; PGPS: Peptidoglycan Polysaccharide

Review of Observational Scoring Techniques in Rodents

Visually scoring an animal’s gait may be the simplest form of gait analysis. Scoring systems often seek to identify “guarding gaits”^{14, 15}, while others report a similar non-specific gait score that characterizes limping or failure to fully apply weight ^16–20. Advantages of scoring include individualized scales relevant to particular disease models, high throughput, and relatively simple data analysis. However, scoring systems are subject to observer bias and typically provide only semi-quantitative gait descriptions.

Recommendations for Gait Scoring in Rodents

Observational scoring is typically performed by placing the animal in an open arena and directly scoring the animal or recording video to be scored later. Due to the rapidity of rodent gait, the human eye is often unable to detect subtle changes without assistance from videography, and these subtle gait compensations may be more likely to associate with diseases like osteoarthritis ^{3, 5, 9, 21}. Hence, we recommend video be used for observational scoring whenever possible. In addition, since scoring is largely an assessment of self-selected behaviors, open arena testing is ideal to capture the animal’s natural gait. Some laboratories have used treadmills and food incentive for these tests, but these factors could mask some gait characteristics ³. Finally, as with any behavioral analysis, investigators should learn proper techniques to handle and acclimate their animals to the experiment.

For studies where gait is not a primary outcome measure, observational scores are a quick, though insensitive, measure of rodent behavior. In arthritis models, these measures could be improved by generating standard scales, similar to the BBB score in spinal cord injury ²². But, this too is inhibited by the broad range of severities observed in arthritis models. As a general recommendation, observational scoring using videography should not use terms such as “mild” or “severe,” as these terms are open to interpretation. Instead, scoring should be distinguished by tangible occurrences, such as the “guarding after noxious compression” definition used by Boettger et al ^{14, 15}. Similarly, without videography, scoring should use a measurable system to categorize animal behavior, such as the ink prints used by Kumar et al¹⁶. Nonetheless, there is a need to refine these scales, but even with standardization, scientists should recognize observational methods will only provide a generic and relatively insensitive measure of rodent gait compensations.

Spatiotemporal Characterization

Terminology

Spatiotemporal gait parameters describe how an animal’s paws move in both space and time, including stride length, step width, duty factor, temporal symmetry, limb phase, and characteristics of the footprint (Figure 1). Spatial parameters are generally familiar variables (Figure 1A). Some gait systems also report variables such as the distance and angle between paws, but scientists should recognize these are not independent measures and can be geometrically derived from stride length, step length, stride length asymmetry, and step width ³. In addition, novice users should remember the difference between step length and stride length; while the left and right step length can differ (spatially asymmetric), the stride length must be the same for all limbs if the animal is using a consistent gait pattern. In addition to these classic spatial descriptors, the orientation and area of each footprint can be analyzed, as well as the relative position of the toes ^23,24.

Figure 1.

(A) Spatial gait parameters are shown for the rat hind limbs. Fore limbs prints were removed for clarity. (B) Temporal gait parameters are described with a Hildebrand plot for quadrupedal rodent gait. Dashed lines indicate a moment in time in the gait cycle, which is depicted by the rat drawing above each line.

Temporal gait parameters are typically less familiar to the non-specialist. The classic parameters of the quadruped temporal gait sequence derive from the seminal work of Milton Hidebrand ² (Figure 1B). These parameters include duty factor, defined as the ratio of the limb stance time and limb stride time^3,11,25(Eq. 1)

$$\mathit{\text{Limb duty factor}}=\frac{\mathit{\text{stance time of limb}}}{\mathit{\text{stride time of limb}}}$$ (1)

To make the nomenclature more obvious to the non-specialist, our laboratory and others have referred to limb duty factor as percentage stance time, but these variables are identical. Temporal symmetry is defined as the time between a right foot-strike and left foot-strike divided by the stride time^3,25 (Eq. 2).

$$\mathit{\text{Temporal symmetry}}=\frac{(\mathit{\text{time of right foot strike}}-\mathit{\text{time of left foot strike}})}{\mathit{\text{stride time}}}$$ (2)

Finally, limb phase is defined as the time between forelimb and hindlimb foot-strike on the same side divided by the stride time (Eq. 3).

$$\mathit{\text{Limb phase}}=\frac{(\mathit{\text{time of left fore foot strike}}-\mathit{\text{time of left hind foot strike}})}{\mathit{\text{stride time}}}$$ (3)

Please note, for a gait sequence to be repeatable, the animal must have approximately the same stride length and stride time on all four limbs.

Review of Spatiotemporal Gait Analysis Techniques for Rodents

A summary of papers utilizing spatiotemporal analyses is provided in Table 3. Early spatial pattern descriptions were analyzed by having the animal walk across an ink pad followed by a sheet of paper and measuring the spatial pattern of the ink prints. Using this technique, rats with antigen induced arthritis had asymmetric spatial patterns indicated by decreased step lengths, and increased paw angles in arthritic rats¹⁵. As discussed below, a major limitation of ink pad methods is the inability to accurately measure animal velocity, a critical covariate for nearly all gait parameters. In addition, the amount of ink is inconsistent between steps and trials, thus paw print areas are highly variable. As such, modern high-speed videography is more likely to provide a robust analysis of spatial parameters.

Table 3.

Review of Spatiotemporal Gait Analysis Used in Rodent Arthritis Models

Species	Joint	Arthritis Model	Gait Method	Parameters Measured	Results	Therapies Tested:	Citation
Mice	Ankle	CFA	CatWalk	Interlimb coordination, stance phase, swing phase, duty factor, stride length, and swing speed	Less regular gait, changes in stance phase duration, swing phase duration, speed, and duty factor. Reduced ratio of right to left hindlimbs for paw pressure, print area, stance phase duration, duty factor, and swing speed.	Indomethacin	11
Rats	Knee	MIA or ACLT	CatWalk	Swing time, swing speed, and duty factor. Velocity as time to cross arena/length of arena	No changes in velocity for any group. MIA: Longer swing phase, slower swing speed, and smaller duty factor in ipsilateral limb. ACLT: No difference in swing phase and swing speed between limbs. Changes in both limbs for swing phase and greater swing speed compared to controls.	Celecoxib	21
Mice	Ankle	LPS	CatWalk	Paw pressure intensity, print area	Ratio of right/left hind paw pressures and the ratio of right/left hind paw areas were lowered after 2 days.	Indomethacin, Minocycline	26
Rats	Knee	Type II Collagenase	CatWalk	Paw print intensity	Reduced paw intensity in collagenase injected rats.	Morphine, Lidocaine, Diclofenac	27
Rats	Knee	MIA	CatWalk	Paw print intensity and area, velocity, stride length, stance time, stride time, swing time	Decreased paw print intensity. No change in other parameters.	None	67
Rat	Knee	ACLT	CatWalk	Velocity, stance duration, swing duration. Calculated a “limb idleness index” from paw print intensity ratios.	Increased LII, target print ratio, and swing duration ratio. Unchanged anchor point ratio. Comparing sham and naïve, there was higher LII, but no difference in ratios.	Buprenorphine	68
Rats	Systemic (rheumatic)	Pristane Injections	CatWalk	Print area, duration of stance phase, regularity index	Decrease in print area, decreased stance phase in 2–3 paws, decreased regularity index. Gait was partially restored at 4 weeks.	None	12
Rats	Knee	MIA	CatWalk	Max contact area, swing speed. Subtracted values from contralateral side	Significant right-left imbalances for max contact areas and swing speed.	Morphine	69
Mice	Knee	DMM	CatWalk	Mean print intensity	No evidence of gait impairment	Morphine, Acetaminophen	70
Mice	Ankle	LPS	CatWalk	Paw pressure intensity, paw print area, and regularity index	Reduced paw pressure, but returned to normal after 4 days. Reduced print areas of all paws –all recovered except the injected limb. Paw pressure ratios and print area ratios decreased, and was most profound at 2 days.	Indomethacin	71
Rats	Paw	CFA	CatWalk	Ipsilateral load percentage (based on area and pressure intensity)	Load reduction on ipsilateral paw	None	72
Rats	Systemic (rheumatic)	Myco- bacterium tuberculosis	CatWalk- like arena	Velocity, stride length, swing time, single stance time, double stance time	Decrease in velocity and stride length, increase in single and double stance time and swing time.	Muscimol, Bicuculline	73
Rats	Knee or Ankle	CFA or Carrageenan	CatWalk- like arena	Paw print intensity, guarding index, regularity index	CFA: Increased guarding index, reduced weight bearing and regularity index. Carrageenan: Increased guarding index, reduced weight bearing and regularity index	Naproxen, Diclofenac, Oxycodone, Ibuprofen	28
Rats	Paw	CFA	CatWalk- like arena	Velocity, hind paw stride length, stride time, swing time, single stance time, dual stance time, and duration of ground contact	Reduced velocity, stride length, single stance time, swing time, and ground contact. Increased dual stance time.	Buprenorphine	29
Rats	Knee	Carrageenan	DigiGait	Swing time, stance/swing ratio, braking time, stance time, % stance/stride, stride length, stride time, % swing/stride, % propulsion/stride, paw area, stance width	Injected limb: Decreased stance/swing ratio, stance, %stance/stride, stride length, paw area, stride. Increased %swing/stride, % propulsion/stride. Contralateral hindlimb: Increased stance/swing ratio, % stance/stride. Decreased % swing/stride, % propulsion/stride. Ipsilateral forelimb: Decreased stance/swing ratio and stride. Contralateral forelimb: decreased stride length and stride	None	38
Mice	Knee	Mechanical Loading	DigiGait	Stance phase, swing phase, stride time, stride length, paw area	Changes only seen in the contralateral limb. Increased stance and stride times, stride length, and paw area	None	74
Mice	Systemic (genetic)	STR/Ort mice	DigiGait	Stance phase, swing phase, stride time, stride length, paw area, brake and propel times, paw angle, symmetry index	Paw area is main parameter associated with OA.	None	39
Mice	Systemic (rheumatic)	CIA	DigiGait	Paw area, paw angle, stride length, stride frequency, stride time, stance time, swing time, braking time, propulsion time	Increased stride frequency and paw area. Decreased stride length and stride time, paw angle, stance time, swing time, braking time, and propulsion time.	None	30
Mice	Knee	Carrageenan	DigiGait	Stride time, stance time, swing time, stride length, stride frequency	Increased swing time. No change in stance time, stride time, stride length, and stride frequency.	Buprenorphine	32
Mice	Knee	TFGβ1 and Running	TreadScan	Stance time, Swing time, Braking time, propulsion time	No change in stride time, stride length, or stride frequency. Increased stance time, propulsion time, and swing time.	Hyaluronan	40
Rats	Systemic (rheumatic)	Myco- bacterium tuberculosis	TreadScan	Stance time, swing time stride length, velocity	Reduced velocity, stride length, increased stance time and swing time	Indomethacin, Aurothiomalate, Chloroquine, D- penicillamine, Sodium Aurothiomalate, Methotrexate	31
Mice	Knee	Carrageenan	TreadScan	Stride time, stance time, swing time, stride length, stride frequency	No change in swing time. Increased stance time, stride time, and stride length. Reduced stride frequency.	Buprenorphine	32
Rats	Knee and Overuse Injury	Strenuous Running	Ink Prints	Stride length and step angles	No difference in stride length, decreased paw angles	None	75
Rats	Knee	CFA	Ink Prints	Limb rotation, stride length, stance width	Increased foot rotation and stance width, reduced stride length	NBQX	76
Rats	Knee	CIA	Ink Prints	Step lengths, angles, and walking speed	Asymmetric gait, decreased step lengths, and increased paw angles in arthritic rats	Enbrel (anti- TNF)	15
Mice	Knee	ACLT and PCL Transection	Ink Prints	Stride length, base of support, paw print area	No change in stride or base of support, smaller foot print after transection	None	37
Mice	Systemic (genetic)	GDF5 deficient mice	Ink Prints	Stride length, base of support	Reduced stride length, no change in base of support	None	77
Rats	Knee	MMT	Custom Gait Arena	Stride length, step width, duty factor, gait symmetry	Asymmetric gait after MMT, imbalanced, different stance time, no change in stride length, step width, and stride frequency	None	9
Mice	Systemic (genetic)	Type IX collagen inactivation	Custom Gait Arena	Velocity, duty factor, stride length, step width, stride frequency, and symmetry	Reduced velocity, increased duty factor, shorter stride lengths, and wider step widths	None	33
Rats	Knee	IL-1β over- expression	Custom Gait Arena	Velocity, stride length/stride frequency, step width, toe-out angle, duty factor, and gait symmetry	Reduced time spent on affected limb and gait symmetry greater than 0.5. Velocity increased, stride lengths increased, and toe- out angles trended up with time.	IL1Ra	10
Rats	Knee	MMT	Custom Gait Arena	Velocity, stance time, swing time, stride time, stride length, and step width. Calculated spatial symmetry, temporal symmetry, duty factor, stance time imbalance, percentage single-limb support.	Asymmetric gait at 2 and 6 weeks post- surgery. Narrow step widths at 2, 4, and 6 weeks. Reduced stride length residuals and 4 and 6 weeks. Temporal asymmetry at 1 and 4 weeks. Stance time imbalance at 1 and 6 weeks.	None	8
Rats	Knee	Carrageenan	Running Wheel	Swing time, swing time ratio	Decreased swing time ratio and swing time	Indomethacin	34
Mice	Systemic (genetic)	Type IX collagen inactivation	Running Wheel, Treadmill, and Custom Gait Arena	Contact time, stride time, duty factor, stride frequency, instantaneous speed	Slower wheel running speed, no difference in duty factor, stride frequency. Overground: No difference in duty factor, stride frequency. Wheel had faster running speeds than over ground, lower duty factor. No difference in stride frequency.	None	78

MIA: Monoiodoacetate Injection; CFA: Complete Freud’s Adjuvant Injection; CIA: Collagen Induced Arthritis (via antigen); LPS: Lipopolysaccharide Injections; ACLT: Anterior Cruciate Ligament Transection; MMT: Medial Meniscus Transection (central tear); DMM: Destabilized Medial Meniscus (at anterior horn)

The CatWalk (Noldus) measures spatiotemporal characteristics along with paw print intensity. In the CatWalk, a light is shined into the side of a glass walkway. When the animal makes contact with the glass, light reflects downward and is captured by a camera beneath the floor (typically recording at 100–150 fps). Advantages of this method include illuminating the portion of the foot touching the floor. The CatWalk calculates 25 parameters based on footprints and 10 parameters based on time, and for arthritis studies, footprint intensities and areas are the most commonly reported parameters^{11, 26–29}. As an example, mice with CFA-induced ankle arthritis had less regular gait with changes in speed and duty factor¹¹.

The DigiGait (Mouse Specifics, Inc.) uses a clear treadmill with a camera underneath to record foot-strikes at 125 fps. DigiGait reports over 50 gait indices, with the parameters most commonly reported including stride length, stance time, swing time, paw area, and braking/propulsion times. On the DigiGait, mice with collagen induced arthritis showed increased stride frequency and paw area and decreased stride length, stride time, paw angle, stance time, and swing time³⁰.

CleverSys offers the GaitScan, which uses a similar approach to the CatWalk, and the TreadScan, which is a treadmill system similar to the DigiGait. A main difference between the TreadScan and DigiGait is the camera (100 fps) positioning; on the Treadscan, a 45° mirror positioned below the treadmill allows a single camera to capture both lateral and ventral views of the animal. Using the TreadScan, rats with adjuvant induced arthritis had a reduction in stride length and increase in stance time and swing time³¹.

Even though DigiGait and TreadScan work on similar principles, these systems do not necessarily achieve the same results. In a rat model of carrageenan-induced arthritis, the 5 parameters relating to the injured foot had opposite results when tested on the DigiGait versus TreadScan³². The authors suggest differences in chamber size, color, lighting, and treadmill belts may explain the variation; however, neither system provided acceptable reproducibility³². This highlights the importance of consistency and is a reminder that small methodological factors can play a large role in rodent gait analysis.

Finally, custom gait analysis systems have been developed by several laboratories. Our lab has constructed an open arena walkway, which allows the animal free mobility with no stressors or rewards. Using a mirror oriented 45° under the floor, we use a high-speed camera (recording at 250 to 1,000 fps) to capture lateral and ventral views of the animal. Having these views allows foot-strike and toe-off (temporal variables) to be measured in the lateral plane and the foot position (spatial variables) to be measured in the ventral plane. Moreover, this platform has allowed us to detect subtle spatiotemporal gait changes, such as spatial and temporal symmetry, duty factor, single-limb support, and temporal shifts of 0.001–0.025 seconds, in multiple rodent arthritis studies^{8–10, 33}.

Recommendations for Spatiotemporal Gait Characterization in Rodents

Many methods are available for rodent spatiotemporal gait analysis, and these systems have considerable advantages over the basic inkpad methods used a few decades ago. Unfortunately, the array of systems and overwhelming number of parameters reported often complicate the gait analysis. We recommend researchers use fundamental descriptors of the spatial pattern – stride length, step length, stride length asymmetry, and step width – and temporal parameters provided by Hildebrand – duty factor, temporal symmetry, stance time imbalance, and limb phase (when forelimbs are included). If available, measures of footprint size and orientation can also be useful. An unfortunate trend has been the reporting of gait variables derived from the above parameters as “new” or “novel”. As an example, swing time ratio (swing time on one foot divided by the swing time on the opposite foot) has been presented as a new gait measure³⁴. However, when examining the Hildebrand plot, a change in swing time is associated with a change in stance time. Thus, swing time ratio measures the same gait change as stance time imbalance. This same shift has also been reported as a change in “single limb support”, a term common in bipedal human gait, but a bit of a misnomer for quadrupedal sequences. Regardless of the preferred nomenclature, the temporal shift is the same – not new.

In addition, investigators must also consider the effects of velocity. Stride length, step length, step width, and limb duty factor are highly correlated to walking speed; thus, accounting for velocity is absolutely essential^{3, 35, 36}. This can be done through statistical models⁸, comparisons to controls^{8–10, 25, 37}, or by controlling velocity with treadmill^{29, 31, 38–40}. However, it should be noted treadmills can mask subtle spatiotemporal changes by inducing stress, and treadmill measurements vary significantly from open arena measurements³. Finally, in open arenas, investigators often provide “home cage” or food incentives. It is not known if these enticements alter rodent gait, since the animal may suppress limb dysfunction and/or pain to reach the reward. As such, we generally recommend open arena testing without an external stimulus, as these methods are most likely to obtain the self-selected gait pattern for each animal.

The camera recording speed can also affect the accuracy of spatiotemporal data. The Nyquist-Shannon rule states an effective sampling frequency should be greater than twice the duration of the fastest factor being measured. The magnitude of temporal shifts will depend on the severity of the arthritic condition. As such, compensations associated with inflammatory arthritis can often be detected at 100 fps. For milder forms of arthritis, our lab has shown frame rates above 200 fps are needed to detect compensatory gait patterns in rodents³.

Finally, but most importantly, understanding how one spatiotemporal gait variable will affect other measures is critical to understanding the pattern. A single spatiotemporal parameter is unlikely to accurately describe subtle gait changes. As an example, both a decrease and increase in duty factor could indicate a gait compensation. For unilateral compensations, the affected limb duty factor decreases while the contralateral limb duty factor increases (limping, imbalanced gait sequence). However, for bilateral compensations, both hindlimb duty factors increase (shuffle-stepping, balanced gait sequence). Both sequences can reduce limb loading; unfortunately, only unilateral compensations are typically considered in arthritis studies. However, we have observed bilateral compensations in a rat model of knee osteoarthritis⁸. Because of the complexity of quadrupedal compensations, it will be difficult to derive a single gait measurement that will capture the myriad of possible compensations. Thus, researchers must carefully consider how the gait pattern has changed both spatially and temporally and how this altered pattern may protect an injured limb.

Kinetic Gait Parameters

Terminology

Kinetics describes forces associated with movement. For rodents, kinetics are largely focused on dynamic weight bearing or occasionally on ground reaction forces (GRF)^{9, 41–43}. For GRFs, the vertical component (z-axis) is the most commonly reported for rodent arthritis models. The anteroposterior force (x-axis) and the mediolateral force (y-axis) may also provide insights to rodent gait; however, the off axis forces (x, y) currently require highly-sensitive, custom platforms^{9, 41, 42}. In addition, the impulse on each axis (area under the force-time curve) is occasionally reported (Figure 2).

Figure 2.

(A) Ground reaction forces are shown for a rat foot with sign conventions. The mediolateral force points toward the midline of the animal, the braking/propulsion force points in the direction of travel, and the vertical force points upward. (B) A standard ground reaction force curve is shown for all three force components.

Review of Kinetic Gait Analysis Techniques in Rodents

A summary of papers utilizing kinetics is provided in Table 4. The Dynamic Weight Bearing Test (Bioseb) uses an instrumented floor and video to calculate the percentage of weight on each leg. An advantage of this system is the animal is free to bear weight normally, unlike the incapacitance test. This test reports parameters including weight, surface area, and time spent on each paw, along with the variability of these measures. The Dynamic Weight Bearing Test has been used in medial meniscus transection and adjuvant induced arthritis studies, finding reduced weight placed on the injured limb in both models^{38, 44, 45}.

Table 4.

Review of Kinetic Gait Analysis Used in Rodent Knee Arthritis Models

Species	Animal Model	Gait Method	Parameters Measured	Results	Therapies Tested:	Citation
Rats	MMT	AMTI Force Plates	Force in x, y, and z. Impulses	Peak vertical force and impulse decreased, propulsive forces and impulse decreased. No change in braking and mediolateral forces.	None	9
Rats	Carrageenan	Dana Load Cells	Vertical Force	Reduced loads after injection. Returned to baseline after 5 days	Morphine	79
Rats	Varus Loading and Compressive Overload	JR3 Load Cell	Force in x, y, and z	Attachment of device: 30% reduced vertical force, reduced A/P force, and M/L shifted. Increased contact time. Overloading: Experimental vs contralateral legs different for all measures except contact time.	None	80
Rats	MIA	Tekscan Walkway	Weight bearing	Reduced weight bearing after MIA.	Naproxen Sodium, Morphine	81
Rats	MIA	Tekscan Walkway	Weight bearing	Reduced weight bearing after MIA. Compensated by increased weight bearing in contralateral hindlimb.	Dexamethasone, Celecoxib, Duloxetine, Naproxen, Morphine, Pregabalin	82
Rats	ACLT	Tekscan Walkway	Paw pressure, impulse	Increased left to right hindlimb average maximum force ratio	Human Synoviocyte Lubricin	46
Rats	MMT	Dynamic Weight Bearing (Bioseb)	Percentage body weight and surface area of each paw	MMT animals shifted weight to forepaws sooner and the forepaw surface area increased sooner. MMT also increased more weight of the contralateral hindlimb.	None	83
Mice	CFA	Dynamic Weight Bearing (Bioseb)	Percentage body weight on hind paws	Significant changes in load distribution were seen as a function of injection concentration.	Indomethacin, Dexamethasone, Morphine Fluorocitric Acid, Minocycline	45

MIA: Monoiodoacetate Injection; CFA: Complete Freud’s Adjuvant Injection; ACLT: Anterior Cruciate Ligament Transection; MMT: Medial Meniscus Transection (central tear)

The Pressure-Sensing Walkway (TekScan) uses an instrumented floor to measure spatiotemporal parameters along with paw pressure and impulse. This walkway has been used to study ACL transection in rats, finding increased hindlimb maximum force ratios⁴⁶. This system allows animals to walk freely. However, users should note the walkway measures foot pressures, which provides a measure of weight born on each limb, but cannot distinguish between the x, y, and z GRFs. Similarly, the CatWalk system (previously described) does not quantify limb forces, but does measure paw print intensity; and this parameter could be considered an indirect measure of limb loading.

Our lab has recently adapted our spatiotemporal arena to simultaneously record kinetic and spatiotemporal data. By instrumenting transparent sections of the floor with 3-axis force plates (Figure 3), 3-component GRFs can be simultaneously determined with spatiotemporal gait parameters. However, this approach requires the animal to make contact with the instrumented floor panels, which markedly increases testing time.

Figure 3.

(A) A craniocaudal view of our laboratory’s gait system is shown with high speed video being recorded from the side (lateral). Force plates are located outside the animal’s path of travel so as to not obstruct view of the feet in the mirror underneath the floor. (B) A lateral camera view of the arena showing force plate positioning and foot print visualization via the mirror.

Recommendations for Kinetic Gait Characterization in Rodents

Again, open arenas promote the most natural gait; however, these arenas may be non-ideal for kinetics due to the significant time associated with waiting for rodents to correctly contact the force panels. Treadmills have helped solve this problem in humans and larger animals^47–50, but creating a force-instrumented treadmill for rodents is difficult due to belt shear forces on the very sensitive and expensive force recording equipment. Moreover, to distinguish between right and left limb forces, human instrumented treadmills (AMTI) have split belts with two force plates, but walking on a split belt treadmill would be difficult to train in rodents. Interestingly, an instrumented running wheel has been developed to measure normal and tangential forces in rodents; however, to our knowledge, this system has not yet been used to study arthritis⁵¹. Running wheels may provide an ideal environment for kinetic data collection, as rodents can use these wheels freely with the probability of a foot-strike on the instrumented rung greatly increased over an open arena.

Because so little is known about rodent GRFs, it is difficult to make strong recommendations on how to analyze these data. Clearly, additional studies and advanced methods are needed, and like spatiotemporal parameters, different compensation strategies may have different effects on kinetic variables (i.e. limping vs. shuffle-stepping). In humans, foot center of pressure shifts laterally with medial knee arthritis⁵², and the free moment increases after knee arthroplasty⁵³. However, it is not yet known whether these data are important or can be measured in rodents. Nonetheless, kinetics tends to be more descriptive of arthritic conditions in larger animals and humans, and as such, these methods may prove more sensitive in rodents.

Kinematic

Terminology

Kinematic parameters describe bony body positions and are commonly defined in terms of three joint angles and three translations (Figure 4). Knee kinematics may also be described as range of motion, which describes the minimum to maximum values for a rotation or translation; however, due to limitations in measuring knee kinematics in rodents, range of motion typically refers to flexion angles in the sagittal plane.

Figure 4.

Sagittal and craniocaudal views of a rodent leg are shown with three translations and three rotations indicated.

Review of Current Methods of Kinematic Gait Analysis in Rodents

Kinematics are commonly studied in humans with skin markers; however, this technology does not scale to rodents due to excessive skin motion artifact^{54, 55}. Nonetheless, skin-based motion tracking systems have been applied to rodents^55–61, but only ankle motion may be considered a reasonably accurate measurement of bone movement⁵⁵. Thus, while clearly valuable, knee, hip, and spine kinematics have not been fully characterized in rodent arthritis models.

Rodent kinematic techniques are developing, with several labs pursuing fluoroscopic techniques to track bone movements⁵⁵. In rats with antigen induced arthritis, sagittal fluoroscopy has detected knee flexion range of motion, with minimum joint angles of 40° in swing and maximum angles of 100° at end of stance¹⁴. Similar studies recorded sagittal plane fluoroscopy in rodents as a proof of concept^{55, 62}, and even these basic fluoroscopic methods greatly improved measurements relative to skin marker tracking. However, current rodent fluoroscopic methods measure 2D flexion angles from sagittal views, and in/out of plane joint positioning may skew these sagittal plane angles. Therefore, biplanar fluoroscopy, which has been used to solve this problem in humans, may also have advantages for rodent kinematics. Biplane fluoroscopy has been used in an evolutionary study of rodents⁶³, but to date, this technology has not been used to study rodent arthritis models.

Recommendations for Kinematic Gait Characterization in Rodents

Again, while open arena testing allows animals to walk naturally, a treadmill may be necessary to collect kinematic data, allowing the animal to stay in the field of view and control radiation doses during the collection of consecutive, repeatable gait cycles. Labs using fluoroscopy may also need to adapt their systems to collect at higher frame rates, since most machines only collect videos at 30 fps. As limb foot-strikes occurring approximately every 0.4 seconds in rats³, this frame rate would only allow 12 images to be collected for a gait cycle.

Rodent fluoroscopy, to date, has largely used single plane, sagittal imaging to study the hindlimbs. Single plane fluoroscopy allows small animals to be positioned closer to the fluoroscopy source to achieve image magnification, and single plane fluoroscopy is more widely available across institutions. However, in other animals, biplane fluoroscopy has greatly improved accuracy and allowed 3D kinematic motions to be better studied. Moreover, metallic markers placed in the femur and tibia may allow more accurate tracking of skeletal motions^64–66, though placing these markers may affect the rodent’s gait. All of these techniques are currently in their infancy, but offer promise to improve our understanding of rodent arthritis models.

Conclusions

The goal of this article was to review current technologies used in rodent gait analysis and provide recommendations for the use of rodent gait analysis in arthritis models. Through this review, we highlighted the relatively large, but current, literature available regarding spatiotemporal gait analysis. For kinetics and kinematics, the literature is more sparse, but viable technologies are clearly being developed. Additionally, though spatiotemporal parameters are the most widely used, several inconsistencies between testing systems and laboratories remain. Thus, rodent gait analysis has significant room for technological advancements in the coming years. When performing rodent gait analysis, it is important to remember that, while rodent gait analysis is a relatively modern concept, the study of quadrupedal gait is not new. Thus, even though most commercially available gait systems report multiple parameters, a balance needs to be found, where critical gait parameters are reported (excluding new ratios or geometric derivatives) while still considering gait changes as an entity. To reach this stage, rodent gait analysis should be advanced to accurately measure spatiotemporal, kinetic, and kinematic parameters, with consistency across laboratories. As rodent gait technologies improve, preclinical models will be better understood and their utility for preclinical arthritis research will increase.

Table 1.

Terminology

Term	Definition
Spatial	Relating to positions in space
Temporal	Relating to time
Kinetic	Relating to the forces associated with motion
Kinematic	Relating to motion
Step Length	Distance from foot strike to subsequent foot strike of the opposite foot
Stride Length	Distance from foot strike to subsequent foot strike of the same foot
Step Width	Distance between the limbs perpendicular to the direction of travel
Duty Factor	Stance time divided by stride time
Temporal Symmetry	Time between a right and left foot-strike divided by stride time
Limb Phase	Time between same side forelimb and hindlimb foot-strikes divided by stride time
fps	Frames per second
GRF	Ground reaction force
Impulse	Area under the force-time curve
Braking/Propulsion Forces	Forces that occur in the direction of travel (also anteroposterior). Typically associated with slowing of the center of mass immediately after foot-strike (braking) and push-off forces propelling the center of mass forward immediately prior to toe-off (propulsion)
Mediolateral Forces	Forces that occur toward the animal’s midline. Typically associated with the stability of the animal and the left-to-right / right-to-left sway of the center of mass during gait