Pros and Cons of Clinically Relevant Methods to Assess Pain in Rodents

Abstract

The primary objective of preclinical pain research is to improve the treatment of pain. Decades of research using pain-evoked tests has revealed much about mechanisms but failed to deliver new treatments. Evoked pain-tests are often limited because they ignore spontaneous pain and motor or disruptive side effects confound interpretation of results. New tests have been developed to focus more closely on clinical goals such as reducing pathological pain and restoring function. The objective of this review is to describe and discuss several of these tests. We focus on: Grimace Scale, Operant Behavior, Wheel Running, Burrowing, Nesting, Home Cage Monitoring, Gait Analysis and Conditioned Place Preference/ Aversion. A brief description of each method is presented along with an analysis of the advantages and limitations. The pros and cons of each test will help researchers identify the assessment tool most appropriate to meet their particular objective to assess pain in rodents. These tests provide another tool to unravel the mechanisms underlying chronic pain and help overcome the translational gap in drug development.

1. Introduction

The ultimate goal of pain research is to develop better treatments to reduce suffering and restore function in chronic pain patients. The problem is that current treatments do not adequately manage chronic pain (Borsook et al.,2014). Barriers to effective treatment include inadequate efficacy of current pharmacological therapies, unpleasant or dangerous side effects that limit dosing, the development of tolerance and dependence with repeated administration, fear of addiction, limited knowledge about how pain is coded, and difficulties assessing the social, emotional, and cognitive aspects of pain. The promise of animal research is to improve understanding of the biological mechanisms underlying pain and the endogenous systems that provide pain relief. Advances in our knowledge of how molecular mechanisms and neural circuits differ between different pain conditions (e.g. evoked vs. spontaneous pain; protective vs. pathological pain; deep tissue vs cutaneous pain) will enhance pain treatment by identifying novel targets. Clinically relevant pain assessment techniques will be especially useful in identifying molecular targets and neural circuits that can be used to guide development of more effective treatments. Using appropriate animal models of pain and thoughtful interpretation of the measures will allow the screening of novel compounds for analgesic efficacy. To date, few treatments targeting pain mechanisms in animals have translated to clinical application in chronic pain patients (Mao, 2012; Mogil et al., 2010a).

There are many reasons for the failure of basic research to produce novel treatments. The endpoint for classical stimulus-evoked pain tests that have been used for decades in rodents von Frey (Frey, 1896), tail flick (D’amour and Smith, 1941), hotplate (O’Callaghan and Holtzman, 1975), and Hargreaves tests (Hargreaves et al., 1988) is the inhibition of a behavioral response following acute application of a noxious stimulus. In many cases, the stimulus is applied to normal, non-pathological skin.Drugs that inhibit these responses in uninjured animals block normal, protective responses to noxious stimuli. Any treatment that prevents a patient from responding to a hot stove or a blister on the toe puts the patient at risk. An ideal analgesic compound will leave pain-evoked responses intact and reduce pathological pain. The goal of the clinician is to reduce pathological pain so the patient can engage in normal life activities, not to block responses to noxious stimuli. This mismatch between the goal of clinicians to restore function and the aim of preclinical studies to inhibit normal pain-evoked responses has limited drug development.

Pain-evoked tests are also problematic because they can’t distinguish between analgesic and motor or other disruptive side effects. An animal’s lack of response to a noxious stimulus could indicate analgesia, paralysis, sedation, or lack of motivation. Some studies deal with this problem by conducting additional tests to screen for side effects (Morgan et al., 2006), but this is time-consuming and most studies do not include such an analysis. A change in preclinical research from using pain-evoked tests to function based tests would facilitate the identification of drugs that inhibit nociception in the absence of disruptive side effects.

The first step in the alignment of animal research and clinical practice was the development of different chronic pain conditions in laboratory animals (Bennett and Xie,1988; Hong and Abbott, 1994;Iadarola et al.,1988; Ness and Gebhart,1988;Schaible and Schmidt, 1985). These studies have advanced animal research from analysis of pain-evoked responses in healthy animals (e.g., hot plate and tail flick tests) to analysis of chronic conditions such as inflammatory and neuropathic pain. The second step in aligning animal research and clinical practice is the use of similar behavioral endpoints. The most common method to assess pain in humans is self-report using visual analog scales, but a number of non-verbal methods that can be used in both humans and animals are also available (Chow et al., 2016;Rabbitts et al., 2014). This review is focused on the pros and cons of behavior measures that can be used to assess clinically relevant pain in both humans and animals.

Our definition of clinical relevance is an animal study that uses a persistent pain condition that occurs in humans and is measured using endpoints that can also be measured in humans. We do not claim, nor do we believe, that only studies that use these criteria are clinically relevant. Clinical relevance depends on the research objective. Pain-evoked tests such as tail withdrawal from hot water may be clinically relevant for studies examining thermal pain mechanisms, but not clinically relevant if the goal is to assess treatments for neuropathic pain. Many pain-evoked tests (e.g., the hot plate & von Frey tests) have been used to assess the analgesic efficacy of drugs to treat pain. These tests have a long history and have contributed much to pain research (Le Bars et al., 2001).However, our more narrow definition of clinical relevance is necessary for the identification of new treatments for chronic pain.

A new generation of clinically relevant pain measures has been developed over the past decade. The tests differ widely, including both pain-evoked and pain-depressed methods. Although some of these tests have been reviewed before (Cobos and Portillo-Salido,2013; Negus et al., 2006;Tappe-Theodor and Kuner,2014), our goal is to describe the advantages and disadvantages of these pain measures as a way to facilitate adoption of the best test for the goals of a particular study.

2. Pain Assessment Tests

2.1. Grimace Scale

The Mouse and Rat Grimace Scale have been widely cited as a clinically relevant method to assess nociception. Although grimacing can also be evoked in aggressive or fearful contexts (Defensor et al.2012), in positive emotional states (Finlayson et al., 2016), and in response to aversive taste experiences (Berridge,2000), pain grimacing is distinguished by the context and associated behaviors (e.g., orbital tightening, ear position). The method to assess pain grimacing consists of examining changes in facial features induced by pain in the same manner that pain is displayed on the face of many species, including humans (Arbour and Gélinas,2014; Costa et al., 2018;Häger et al.,2017; Hampshire and Robertson, 2015). Video images of the animal’s face are captured and scored using a 0 (not present) to 2 (obvious) rating scale for each facial feature. These ratings are averaged across images taken at different times to produce a pain score. The images can be cropped to focus specifically on the face so the observer is blind to the treatment condition.

There are a number of advantages to using the Grimace Scale.Grimacing is a recognized expression of pain across a wide range of species. Although there are species-specific differences, the test can be used with non-verbal humans, pets, and laboratory animals. The method is relatively simple and requires no specialized equipment other than a video camera and frame capture software. The animal does not need to be trained and images can be assessed in the home cage which limits confound from non-pain factors (i.e., novelty; stress). Finally, the Grimace Scale can be used to assess both evoked- and spontaneous pain.

The primary limitation of the Grimace Scale is the mismatch between pain duration and expression of a pain face. Induction of hind paw inflammation produces allodynia to pressure that persists for over two weeks, but the Grimace Scale score returns to baseline between 2 and 7 days (Sotocinal et al.,2011). This difference may reflect the different time course of evoked andspontaneous pain, but prevents the use of the Grimace Scale for long lasting pain conditions such as that caused by neuropathy (Langford et al., 2010). In such a case, the lack of grimacing does not mean that the animal is pain free, just that the pain no longer elicits a grimace. Moreover, the grimace scale is useful in monitoring analgesic effects during the early phase of neuropathic pain in rats (Philips et al., 2017).

Although it is technically possible to collect video images continuously, data analysis is limited by inadequate images and the time needed to evaluate each image. In addition, observers must be trained to use the scoring system. Such training provides surprisingly high inter-observer reliability (r = 0.9) (Sotocinal et al.,2011) given the subjective scoring system. Recently, face recognition software has been adapted to reduce the time and tedium of conducting grimace analysis (Tuttle et al., 2018). Facial features characteristic of pain are evident 8% of the time in pain free rats, and the incidence of these spontaneous responses has been shown to vary across sex (male > female) and mouse strain (Miller and Leach, 2015). Mean pain score rarely exceeds 1 on the 2-point scale, requiring large sample sizes or significant levels of pain to produce statistically significant differences. Administration of morphine reduces this difference as would be expected of an analgesic (Sotocinal et al.2011), but these data are confounded by sedative and motor effects, as in all measures that use pain-evoked behavioral responses.

2.2. Operant Behavior

A decrease in behavior is a diagnostic criterion for pain in both humans and other animals. In humans this can be manifested as a reduction in exercise, work, or socializing. A similar assessment can be conducted in animals by measuring pain-depressed operant behaviors. For example, Martin and colleagues showed a reduction in operant responding for food following surgical pain in rats (Martin et al., 2004). Operant responding for intracranial self-stimulation (ICSS) (Olds and Milner, 1954) also has been used to assess the effects of pain (Ewan and Martin, 2014;Lazenka et al., 2018;Leitl and Negus, 2016). Rats are implanted with a stimulating electrode into the medial forebrain bundle and trained to press a bar for brain stimulation. Once stable levels of responding are evident across days, a noxious stimulus is applied and changes in the response rate are measured.

The primary advantage of assessing pain-depressed behaviors such as operant responding is the ability to distinguish the analgesic and sedative/motor effects of treatments. Drugs that produce analgesia restore pain-depressed responding, whereas drugs that produce disruptive side effects (e.g., motor, sedative, motivation) reduce operant responding. High levels of stable responding occur with operant responding for food or ICSS, providing an ideal background to detect pain-induced depression of behavior. Moreover, the level of responding can be manipulated by altering the reward schedule or adjusting stimulation parameters. Adjusting these parameters can be used to assess the potency of a particular pain condition to disrupt behavior or to reduce variability between animals. Operant responding also produces a clear and objective measure of behavior.

Although operant conditioning experiments initiated the expansion of preclinical pain studies to focus on the motivational consequences of pain, few laboratories have adopted them as a pain assessment tool. Almost all of the research on pain depressed ICSS are from two labs (Ewan and Martin, 2014; Negus, 2013). From a practical standpoint, ICSS studies are limited by the equipment needed for operant responding, the time to implant and train animals to respond, and the potential for surgical pain to confound subsequent pain assessment.Once implanted and trained, animals can be tested repeatedly to assess changes in behavior as pain evolves day to day. However, test sessions are typically limited to 30 −60 minutes.

The use of ICSS as a pain assessment tool is also limited by the complex relationship between operant responding for rewarding stimuli and pain. Hindpaw inflammation and neuropathy are two well-established pain conditions in animals that have little or no effect on ICSS (Ewan and Martin, 2014Lazenka et al., 2018;Leitl and Negus, 2016). The decrease in ICSS caused by noxious stimuli such as administration of intraperitoneal acid, incisional pain, and formalin injection into the hindpaw (Lazenka et al.,2018;Leitl and Negus, 2016) may be caused by competing pain-evoked behaviors such as stretching or licking the hindpaw. Likewise, it is difficult to determine whether analgesics have a direct effect on pain or alter operant responding by modulating reward pathways. The ability of nicotine, opioids, and dopamine receptor agonists to reverse the depression of ICSS caused by intraperitoneal administration of acid (Ewan and Martin, 2011; Freitas et al., 2015;Lazenka and Negus, 2017; Miller et al., 2015; Miller and Leach, 2015) could be caused by inhibition of pain or enhancement of the rewarding effects of ICSS. The good news is that these problems can be mitigated by the use of appropriate control groups (e.g. non-injured animals). The real benefit of these studies has been to raise awareness of the value of pain depressed behavioral tests. The subsequently described pain depressed behavioral tests (wheel running, burrowing, home cage monitoring) inspired by these ICSS experiments have all of the advantages and few of the disadvantages.

2.3. Wheel Running

Wheel running is a natural rodent behavior that has been observed in rats and mice in the wild when a wheel is made available (Meijer and Robbers, 2014). Wheel running can be used as both an independent (e.g., exercise) and dependent (e.g., activity) variable in the laboratory (Sherwin, 1998). Most rats and mice will run by simply providing access to a wheel either in the home cage or by moving the animal to a running wheel apparatus. In 2004, Clark and colleagues used wheel running as a dependent variable to show that oxymorphone could restore activity following splenectomy in mice (Clark et al., 2004). Since then, a range of other pain conditions have been shown to depress wheel running in rats and mice. These include hindpaw inflammation, arthritis, migraine-like pain, neuropathic pain, surgical pain, and more (Cattaruzza et al., 2013;Kandasamy et al., 2016, 2017b; Kendall et al., 2016;Pitzer et al., 2016a, 2016b; Stevenson et al., 2011; Tang et al., 2016; Whitehead et al., 2017).

Pain reduces voluntary activity in humans (Mannix et al., 2016; Rabbitts et al., 2014;Stang and Osterhaus, 1993) and is included as a diagnostic criterion for some chronic pain conditions (Headache Classification Committee of the International Headache Society, 2013). Wheel running is an easy to use and objective method to assess the impact of pain on animals. The animal is simply placed in a cage with a running wheel. Each wheel revolution is digitally counted and the time stamp stored by a computer. The data are collected in a completely objective manner. Home cage wheel running has the added advantages of preventing stress from moving the rodent to a test chamber and data can be collected continuously. Given that rats and mice are nocturnal, most running occurs during the dark phase. Both the magnitude and duration of pain- or drug–induced recovery can be assessed by measuring changes in wheel running over time (minutes to days).

Unlike pain evoked tests (e.g., hot plate; von Frey tests) in which the motor or sedative effects of treatments can confound assessment of antinociception, treatments that produce disruptive side effects will not restore wheel running. This feature makes pain-depressed wheel running especially useful in screening drugs because only treatments that produce antinociception in the absence of disruptive side effects will restore wheel running. A surprisingly low dose of morphine (1 mg/kg) restores wheel running depressed by hindpaw inflammation. Higher doses, such as those to block pain-evoked responses (doses > 3 mg/kg in most rat studies) depress wheel running because of the side effects (Kandasamy et al.,2017a). Finally, depression of home cage wheel running is an especially useful method to assess spontaneous pain as occurs in migraine (Kandasamy et al., 2017b).

One of the limitations of wheel running is that stable levels of running may require up to 17 days of exposure (Stevenson et al., 2011). Running levels are sufficiently high after 7 days of wheel exposure that most experiments can be initiated within a week (Kandasamy et al., 2016; Stevenson et al., 2011), but activity levels will tend to drift upward with short habituation periods. A more serious problem is managing variability. Some rats do not run enough to see meaningful changes in activity so exclusion criteria are required to remove animals that do not have reasonable baseline levels of running. A recent study removed approximately 15% of the rats for insufficient levels of baseline activity (Kandasamy et al., 2016). At the other extreme, some rodents run up to 10 km/day. This between-animal variability can be controlled by converting the number of wheel revolutions to a percent of each animal’s baseline activity. Variability can be especially great when comparing male and female rats. In some strains, female animals run approximately twice as much as males. Wheel running also has been shown to vary across the estrous cycle (Loram et al., 2007) and across rat and mouse strains (Clark et al., 2011; Gordon et al., 2016), although our experience indicates mice have more consistent levels of running between animals (Pitzer et al., 2016a, 2016b). Variability is also high because animals get on and off the wheel throughout the night. This variability can be reduced by compiling data over long time periods (e.g., 3 hour blocks or daily).

Given that over 95% of wheel running occurs during the dark phase of the day/night cycle (Kandasamy et al., 2016;Pitzer et al., 2016b), the impact of pain and drug treatments are only evident during the dark phase. Although it is possible that pain conditions that disrupt sleep could increase running during the light phase, no obvious examples of this have been reported to date. Variability is also introduced by the transient increase in wheel running that occurs any time the animal is removed and returned to its cage. We address this problem by only handling animals once a day and beginning data collection 10 min after the animal is returned to its cage.

The most significant limitation of home cage wheel running is that it is an indirect measure of pain. Any disruptive stimulus, such as illness (Taraborrelli et al., 2011) the side effects of high drug doses (Kandasamy et al., 2017a) or opioid withdrawal (Kandasamy et al., 2017c) depresses wheel running. In some conditions, such as chemotherapy-induced neuropathic pain, it is difficult to distinguish whether pain or illness is the cause of depressed wheel running. Stress, motivation and other factors may also influence wheel running. For example, housing animals individually can induce stress and this has been shown to attenuate the depression of wheel running caused by neuropathic pain in mice (Pitzer et al., 2016b). Technological solutions for group housed animals include the use of video or telemetry to identify specific animals.

A final problem is the interaction between exercise and pain. Prolonged access to a running wheel facilitates recovery from pain (Brito et al., 2017; Grace et al.,2016; Sabharwal et al., 2016). This interaction complicates analysis of wheel running as a measurement of pain, but enhances the clinical validity of these preclinical studies. Exercise is an effective pain treatment in humans and something physicians encourage to facilitate recovery. Pain is probably exacerbated in most animal studies because movement is limited by housing animals in small cages. Home cage wheel running is consistent with medical practice by allowing animals to voluntarily increase their activity as they feel better which in turn facilitates recovery.

2.4. Burrowing & Nesting

Nesting and burrowing are natural rodent behaviors used to assess wellbeing, including the assessment of pain (Jirkof,2014). Nesting is assessed by measuring the time or quality of nest building following introduction of nesting materials in a rodent’s cage. Burrowing is assessed by measuring the time or amount of material (e.g., food, bedding) removed from a tube place in a rodent’s cage. These tests produce consistent results across labs (Wodarski et al., 2016) and have been shown to be sensitive to surgical, neuropathic, and inflammatory pain (Andrews et al., 2011; Arras, 2007; Huang et al., 2013;Jirkof et al., 2012, 2013; Lau et al., 2013b; Van Loo et al., 2007; Rutten et al., 2014b, 2014a).

A major advantage of these tests is that they are conducted in the home cage, avoiding stress and other confounds caused by testing in a novel environment. The tests can be repeated daily or more frequently by removing and re-introducing nesting or burrowing materials. Testing typically occurs during the active dark phase allowing researchers to observe the result during the light phase when the animal is inactive.

One problem that occurs with all pain-depressed behavioral tests (e.g., ICSS, wheel running) is that disruption of nesting and burrowing are not specific to pain. Any stimulus that disrupts well-being, including disruption of memory (Deacon et al., 2002, 2008) can decrease nesting and burrowing. This problem requires the use of appropriate control groups as a comparison. Nesting appears to be an innate behavior for most rats and mice, although this test cannot be used in species (e.g., Egypt spiny mice) that do not build nests naturally (Deacon, 2009). Burrowing has also been shown to be enhanced with practice or when other animals are present (Deacon et al., 2001; McLinden et al., 2012). Sex, social situation, and room temperature also influence nesting with females housed together in a cold environment producing higher nesting scores than females housed singly or compared to males (Gaskill et al., 2013; Van Loo et al., 2007). Although these factors can influence the results of an experiment, they are easy to manage by experimental designs that include control groups and consistent testing procedures.

A bigger problem is how to assess nest building. The time to initiate or build a nest provides an objective method to assess nest building, but it fails to assess the quality of the nest. The problem with a rating scale to assess nest quality is that it is subjective. Time can also be used to assess changes in burrowing, although the more common approach is to measure the amount of material removed from the tube. Both nesting and burrowing occur primarily during the dark phase of the light cycle, but animals will start these behaviors immediately after exposure to the material (Jirkof et al., 2010, 2013). It is therefore important to keep the duration and time of day consistent. Repeated testing is possible by introducing new materials into the cage and measuring the latency for the animal to manipulate the new material.

2.5. Home Cage Monitoring

Rodents, like humans, show alterations in their daily behavior when they are ill. Sickness behaviors comprise decreased food consumption, drinking, grooming and increased sleeping (Aubert, 1999; Hart, 1988). Changes in body weight and fur appearance can be assessed also. Many different types of home cage monitoring systems are commercially available (Table 1). These systems use infrared light beams, video, or weight/vibration sensors to measure overall and specific types of locomotion (e.g., climbing, grooming, eating, drinking, rearing, and scratching). The use of home cage monitoring as a method to assess pain in rodents is currently limited, and of the few studies that have been published, none show long term behavioral alterations (Bree et al., 2016; Inglis et al., 2008; Leach et al., 2012; Pitzer et al., 2016a, 2016b; Urban et al., 2011). Acute behavioral alterations have been shown following postsurgical pain (Bree et al., 2016; Leach et al., 2012), neuropathic pain (Pitzer et al., 2016b; Urban et al., 2011), and osteoarthritis (Inglis et al., 2008). The effect of hindpaw inflammation on home cage parameters are mixed (Pitzer et al., 2016a; Urban et al., 2011). There are clear advantages to assessing pain using home cage monitoring. Analysis of behavioral alterations is performed in a home cage under standard housing conditions (depending on the system being used) which reduces stress (Tecott and Nestler, 2004). There is no limit to the duration of analysis, and behavior can be assessed over the entire circadian rhythm. Rodents are mostly active during the dark phase and potential pain-related behaviors can be automatically assessed during this time. Moreover, it is an observer independent and objective method. Commercially available systems range from simple analysis of movement in three directions to fully automated analysis of up to 20 behavioral measures.

Table 1:

Commercial available homecage behavior analysis systems

Name	Company	Features
CLAMS-HC Or Opto-M4 (animal activity meter)	Columbus instruments (USA)	individual housing IR photocell based automated assessment of activity parameters (x, y, z direction), feeding, drinking, temperature and heart rate (potentially combined with wheel running)
Micromax	AccuScan Instruments (USA)	individual housing IR photocell based automated assessment of activity parameters (x, y, z direction
SmartFrame Home Cage system	Kinder-scientific (USA)	individual housing IR photocell based automated assessment of activity parameters (x, y, z direction
HomeCageScan	CleverSys (USA)	individual housing automated video analysis automatic detection of more than 20 behavioral measures requires good lighting conditions or IR-switchable camera or red-light for night
Phenotyper	Noldus (Netherlands)	individual housing automated video analysis (Ethovision tracking, Observer scoring) using IR-sensitive camera automatic detection of more than 10 behavioral measures
Phenocube	Psycho-genics (USA)	grouped housing RFID based, analysis of multiple animals per cage, social interaction automatic detection of several behavioral measures only customized fee-for-service testing
LABORAS	Metris (Netherlands)	individual housing fully automated movement detection via specific platform automatic detection of more than 10 behavioral measures
Intellicage	TSE (Germany)	grouped housing RFID based, analysis of multiple animals per cage, social interaction automatic detection of different behavioral and cognitive measures, learning paradigms
ActualHCA	Actual analytics (UK)	grouped housing RFID and video based, analysis of multiple animals per cage, social interaction automatic detection of several behavioral measures
Activity monitor	PHENOSYS (Germany)	grouped housing RFID based, analysis of multiple animals per cage Automatic measurement of different locomotor behaviors

Most home cage monitor systems are limited by having to house the animal individually. Stress is enhanced by limited enrichment in the environment to prevent obscuring data collection. The use of telemetry (Radio-frequency identification) allows the tracking of individual animals under group housing conditions, but only a few of the commercially available systems incorporate this technology (e.g.,Phenocube, Intellicage, ActualHCA, PHENOSYS). Depending on the endpoints and specific resolution, data from individual animals housed together can be collected.

Video monitoring of home cage behavior is limited by manual analysis of behavior. Fortunately, systems that use infrared light beams or weight/ vibration sensors to measure behavior allow for automation. Although animals can be maintained in these cages for prolonged periods of time, high levels of exploratory behavior when the animal is first placed in the cage can confound the data (Bains et al., 2016; Pitzer et al., 2016a; Urban et al., 2011). Moreover, depending on the pain-model and analysis system, the number of animals that can be tested is a significant limitation. Most home cage monitoring systems are expensive and require a significant amount of space.

The effects of pain on standard home cage behaviors tend to be small. Quadrupeds seem to more easily compensate for specific deficits (e.g., back pain, injury to a single limb) in home cage activity than a biped. The automatic systems are not sensitive enough to assess behavioral alterations like spontaneous foot lifting (Djouhri et al., 2006), an accepted indication of spontaneous pain (Lolignier et al., 2011). The recent development of a behavioral spectrometer allows very sensitive evaluation of specific behaviors (e.g. specific body-oriented grooming) (Brodkin et al., 2014). Integration of these sensitive measures with standard home cage monitoring would greatly enhance pain assessment. A final problem is how to interpret the data. Is a reduction of climbing activity pain-mediated or based on reduced motivation to escape?

2.6. Gait Analysis

The use of gait analysis to study locomotor abnormalities has a long history. Early studies involved dipping a rodent’s hindpaws in ink and manually analyzing footprints as the animal walked across white paper. Analysis was enhanced with the development of a treadmill with a high speed camera (Heglund et al.,1974). Application of gait analysis to pain conditions took longer to develop. Most pain studies focused on models of arthritis (Adams et al., 2016; Clarke et al., 1997) and expanded to hindpaw inflammation and neuropathic pain later (Coulthard et al., 2002; Lau et al., 2013a;Pitzer et al., 2016a, 2016b;Vrinten and Hamers,2003). Both commercial and custom gait analysis systems have been developed that allow analysis of static, dynamic, and kinetic gait parameters. In addition, there are obvious gait abnormalities that are identifiable by the naked eye and can be scored easily. Among these are different categories comprising guarding of a paw or limb, no usage of a limb, no movement at all, visible limping or leg dragging (Lakes and Allen, 2016). Although the specific procedure varies depending on the system used (Table 2), the procedure typically includes one or two acclimatization sessions followed by up to three test runs. Each test consists of placing the animal in the start box and allowing it to walk across an enclosed space while recording each step. Changes in static and kinetic gait can be analyzed.

Table 2:

List of gait analysis systems

Name	Company		Features
Footprint test/ Ink prints	home-made		hindpaws are painted with ink and the animal is allowed to walk across a paper analysis of stride length, limb rotation harbors stress and is work intensive manual analysis
Static weight bearing -incapacitance	Panlab (USA); iitc (USA); Bioseb		the rodent is placed into a restrained holder; hindpaws rest on two separate sensor plates automatic quantification of weight distribution
test	(France), Linton Instr. (UK),	no walking parameters assessable
Dynamic Weight Bearing Test	Bioseb (France)	rodents are placed on a small arena floor pad sensor equipped with multiple pressure transducers (16/cm2), combined with a video camera semiautomatic analysis of static parameters no walking parameters
Kinetic Weight Bearing Test	Bioseb (France)	rodents have to cross a floor walkway (voluntary walking) equipped with multiple pressure transducers, combined with a camera on top automatic quantification of approx. 20 different parameters
CatWalk	Noldus (Nether-lands)	rodents have to cross an over ground glass floor walkway (voluntary walking), illuminated by internally reflected light, reflected light is captured by a high frame camera beneath the floor automatic quantification of more than 40 different parameters
DigiGait	Mouse-Specifics (USA)	rodents walk on a possibly motorized transparent treadmill system (voluntary or forced walking) that detects paw position via image registration by a high frame camera from below automatic quantification of over 50 different parameters
TreadScan or GaitScan	CleverSys (USA)	rodents pass a transparent belt treadmill (TreadScan; forced walking) or a clear free-walk runway (GaitScan; voluntary walking), a high speed camera from below assesses the walk automatic quantification of approx. 20 different parameters
GAIT® analyzer 21	Noveltec (Japan)	rodents are placed into a transparent running wheel (forced wheel running), a high speed camera captures foot bottom pictures of the hindlimbs automatic analysis of approx. 10 dynamic hindpaw parameters
Pressure-Sensing Walkway	TekScan (USA)	rodents have to pass a walkway with pressure-sensitive sensors (15.5/cm2) within the floor automated calculation of different gait parameters

There are a number of advantages of using gait analysis to assess pain. Gait analysis can be used across a wide range of species, and the animal literature is supported by a large number of human studies examining the effect of pain on gait (Constantinou et al., 2014; Phillips and McClinton, 2017;Toosizadeh et al., 2015) The measurements are performed in freely moving animals. Commercially available systems allow data to be collected and analyzed objectively. Assessment is easy, quick, and requires no training other than one or two acclimatization sessions. Repeated testing can be carried out allowing before and after treatment assessments and analysis of changes in gait over time. Gait analysis is particularly useful in evaluating potential treatments because only drugs that improve movement will pass screening. Drugs that produce sedative or motor effects will not restore normal walking and would not be recommended as an effective treatment because of these side effects. Data tend to be reproducible across trials and animals within a specific condition.

Gait analysis is a specialized pain test limited to pain conditions that affect movement either by directly (e.g., arthritis) or indirectly (e.g., back pain) impacting the limbs. It cannot be used for pain conditions that don’t impact movement or conditions that cause a complete inhibition of movement. Animals that do not move because of pain or lack of motivation cannot be tested. Given that many factors can alter gait, interpretation of data can be difficult. Variations in gait between species, strains, age, sex of the animal, and weight of the subject limits comparisons across studies and requires well-matched control groups within a study. Some pain conditions and treatments can cause a decrease in body weight that leads to a change in gait. Moreover, pain conditions such as arthritis and neuropathy cause both pain and motor effects, either of which can alter gait. The lack of a correlation between the time course for neuropathy-induced mechanical hypersensitivity and change in gait, in addition to the lack of recovery of normal gait following treatment with standard analgesics would indicate a motor problem (Lau et al., 2013a; Mogil et al., 2010b) (Shepherd and Mohapatra, 2018). Although neuropathic pain also appears to alter gait, some parameter changes are transient (Pitzer et al., 2016b).

Comparison between studies is further complicated by the different parameters assessed by different analysis systems (see Table 2). A common factor is that these systems are expensive. Few studies directly compare different gait systems. Pitzer et al. compared the dynamic weight bearing system and the Catwalk in CFA and SNI models and found comparable results (Pitzer et al., 2016a, 2016b), whereas the DigiGait and TreadScan systems were not consistent in a rat model of carrageenan-induced arthritis (Dorman et al., 2014). The authors discuss differences in chamber size, color, lighting, and treadmill belts as potential underlying factors (Dorman et al., 2014).

2.7. Conditioned Place Aversion and Preference

Conditioned Place Avoidance (CPA) and Conditioned Place Preference (CPP) are well-known methods to assess motivational components of pain and pain relief. CPA builds on the protective functions of pain to motivate escape and avoid harm. Pairing a painful experience with a distinct context results in subsequent avoidance of the same contextual cues (Johansen et al., 2001; Johansen and Fields, 2004). CPP is the opposite of CPA and is based on the assumption that pain relief is rewarding. Pairing a rewarding experience with a distinctive environment will increase the time spent in that environment (Navratilova and Porreca, 2014; Navratilova et al., 2015, 2016)The most common procedure uses two conditioning chambers distinguished by visual, textural and sometimes odor cues. Some scientists use a neutral chamber that the animal must pass through to get to the conditioning chamber on either side. CPA occurs by pairing a painful experience with a specific chamber. CPP occurs by pairing pain relief with a specific chamber. Depending on the number of conditioning trials, a single test takes 3 to 9 days to complete (Johansen et al., 2001; King et al., 2009). Testing begins with habituation to the CPA/CPP chamber for 1 to 3 days prior to conditioning to determine a potential side preference. This is followed by one to four conditioning sessions and a similar number of placebo trials. Placebo trials can be performed on the same day as the conditioning trials or on alternate days. The test session is conducted 24 hours after the final conditioning session in the absence of any manipulation by measuring the amount of time the animal spends in each chamber. A decrease in time spent in the conditioning chamber indicates CPA, whereas an increase in time indicates CPP.

Treatments that alleviate pain induce CPP across a range of pain conditions such as incision, inflammation, post-operative pain, nerve injury, osteoarthritis, cancer pain, spinal cord injury, and cephalic pain (Davoody et al., 2011; De Felice et al., 2010, 2011;Havelin et al., 2017; Hung et al., 2015; King et al., 2011; Liu et al., 2011;Okun et al., 2011, 2012, Park et al., 2016, 2013;Remeniuk et al., 2015;Sufka, 1994;Xie et al., 2014; Yang et al., 2014). Drugs that are not rewarding in the absence of pain become rewarding in the presence of chronic pain. Treatments that do not directly activate the reward pathway produce CPP in animals with chronic pain. For example, peripheral nerve block, administration of spinal ω-conotoxin or clonidine, and RVM lidocaine induce CPP in animals in pain but not in control animals (Havelin et al.,2017;King et al., 2009, 2011,Okun et al., 2011, 2012;Remeniuk et al., 2015). Relief from the aversive component of chronic pain motivates the animal to increase time in the treatment-paired chamber.

The ability to measure the affective component of pain and pain relief is a major advantage of using CPA and CPP to study pain. Another important advantage of CPP is that it measures the non-evoked component of chronic pain, often referred to as the spontaneous or ongoing component of pain. Such spontaneous pain appears to be particularly important in motivating patients to seek medical care. CPP has been reliably demonstrated in response to drugs that are effective in alleviating spontaneous/ongoing pain clinically (e.g. spinal clonidine, ω-conotoxin, systemic gabapentin, duloxetine), but not drugs that fail to alleviate spontaneous pain (King et al., 2009). The focus on spontaneous/ongoing pain allows for exploration of mechanistic differences between spontaneous and evoked pain states (Havelin et al., 2017; King et al., 2011;Okun et al., 2011;Remeniuk et al., 2015). Another advantage of CPP is that because assessment of the analgesic effects of treatment occurs 24 hours after the last drug treatment, CPP is not confounded by undesirable side effects such as sedation or inhibition of movement. In addition to assessment of pharmacological agents, CPP and CPA have been used with optogenetic and chemogenetic tools to examine circuits that underlie pain (Cai et al., 2018; Hirschberg et al., 2017;Seo et al., 2016; Tan et al., 2017; Zhang et al., 2017). Such efforts will improve our understanding of how neural circuits differ between processing different components of pain (affective vs somatosensory) and across different pain conditions (e.g. evoked vs. spontaneous pain).

One disadvantage of CPP/CPA is that it is a complicated and time-consuming process. Each animal requires approximately 30 minutes of conditioning/testing each day, and habituation, conditioning, and test trials can take up to 9 days. The duration of each conditioning trial depends on the duration of the pain stimulus or drug treatment to maximize association between these stimuli. Long-term or chronic pain states are not suitable for inducing conditioned place aversion as they persist outside the context and are therefore not specific to the distinctive environment. Most studies use multiple pairings to establish the avoidance (Johansen et al., 2001; Johansen and Fields, 2004). However, a single pairing can be sufficient to induce CPA in some pain models such as movement-induced breakthrough pain (Havelin et al., 2017).

It is essential to consider the route of administration and pharmacokinetics of drug treatments. Important aspects of treatment include onset and time-course of effect. If the onset is slow and the peak occurs after the pairing period ends, this may diminish conditioning. A rapid onset along with a time-course of effect that persists throughout the conditioning period is optimal. Given that there may be important differences between evoked and spontaneous pain (Okun et al., 2011;Xie et al., 2014), the optimal conditioning time may not be evident. Analysis of the time course for analgesia is costly because it requires the conditioning of different animals at different time points.

Although variability is a natural part of any behavioral experiment, the high variability and small effects that occur with CPA/CPP require large sample sizes to provide sufficient power for statistical analysis. Previous studies use sample sizes greater than 10 animals per group. Large sample sizes can be a particular challenge when using transgenic animals. Most studies use an unbiased design in which the two conditioning chambers are calibrated so that pre-conditioning time spent in the chambers is comparable. To diminish variability, properly calibrating the chambers so that animals spend roughly equivalent time in each of the chambers prior to routine use is critical. In addition, consistent exclusion criteria can be used to reduce variability from animals displaying a strong bias (e.g. King et al., 2009). This must be done consistently and applied prior to conditioning.

Drugs used to treat pain can have a wide range of side effects. Given that CPA and CPP require conditioning, drugs that are rewarding (e.g., morphine) or aversive (e.g., kappa opioid receptor agonists) make data interpretation difficult. Chronic pain can also increase aversion to an acute pain (Zhang et al., 2017). The inclusion of uninjured control groups is essential to determine the magnitude of reward or aversion independent of pain relief. Likewise, factors that influence memory such as outside environmental influences (e.g. noise, cage changes) or drugs that impact learning or memory can obscure analgesic effects. Appropriate controls are critical for adequate interpretation of the data in these cases. Despite the many difficulties described above, CPP/CPA experiments have played a key role in understanding the affective component of pain.

2.8. Other Tests and Issues

All of the tests described above have been promoted as clinically relevant methods to assess nociception in animals. Many of these tests (e.g. ICSS, CPP) have been used for years, but only recently applied as a pain assessment tool. Other tests like sucrose preference, a method to measure anhedonia, or ultrasonic vocalizations have also been used to assess pain (Neugebauer et al., 2007; Wang et al., 2011) Pain often produces comorbid symptoms such as anxiety and depression that can influence pain behaviors or be assessed specifically. Anxiety-like or depression-like behaviors have been shown to vary in a time dependent manner in models of neuropathic pain (Sellmeijer et al., 2018; Seminowicz et al., 2009). Some of the preclinical tests described above may be especially sensitive to housing conditions and handling stress that affect emotions. For example, social isolation has been shown to prevent SNI-induced depression of wheel running in C57Bl6 mice (Pitzer et al., 2016b). Other factors that might influence test results include strain and sex. Although these factors have been assessed (Kandasamy et al., 2016; Miller et al., 2015), more research is needed to determine how strain and sex influence data on each test.

3. Conclusion

The key to successful research is using tests that match the goals of the study: For example, gait analysis to understand the impact of arthritis on movement, reversal of pain-depressed wheel running for drug development, and CPP to assess the affective component of pain. We do not claim that the tests described in this review are better than other tests, but they do mimic clinical features of pain in unique and important ways. Our goal in describing problems with each test is not to discourage use, but to make the reader aware of the limitations. The appropriate test depends on the pain condition and the specific methods used (Sheahan et al., 2017). All nociceptive tests have advantages and disadvantages. Advocating for one test over other tests as some have done (Vierck and Yezierski, 2015) fails to recognize that the suitability of a test depends on the goal of the study. If the goal is to understand the processing of acute noxious thermal stimuli, then the hot plate and tail flick tests are appropriate tools. If the goal is to test the efficacy of new analgesic compounds, then clinically relevant pain conditions and assessment techniques as described here are needed. If the goal is a better understanding of differences between pain conditions (e.g. evoked vs spontaneous pain), a combination of measures is appropriate. Moreover, combining tests provides a more complete understanding of the many facets of pain. Although the efficacy of some of these tests has been questioned (Sheahan et al., 2017), our hope in describing the advantages and limitations of each test is to help researchers select the most appropriate test. There is no doubt that pain research will be enhanced by a wider use of the tests presented in this review.

Highlights.

• Successful translational preclinical research needs meaningful behavioral tests

• New tests have been developed to focus more closely on clinical goals such as reducing abnormal pain and restoring function (Grimace Scale, Operant Behavior, Wheel Running, Burrowing, Nesting, Home Cage Monitoring, Gait Analysis and Conditioned Place Preference/ Aversion) and mimic clinical features of pain in unique and important ways

• Advantages and limitations of these test are highlighted and enable researchers identify appropriate tests meeting their particular objective