A review of pain behaviour assessment scales in horses

Methods to assess clinical pain

Pain can only be treated after it has been detected, but the decision to provide analgesic treatment is not only driven by whether pain is present or not. Identifying the type and intensity of pain are key to guiding the choice of the best analgesic therapy for a horse experiencing pain. To do this, the pain needs to be scored and allocated a ‘number’. After that, the success or failure of a given analgesic can be checked by determining if, following analgesia, the score has reduced below the cut-off point of the pain scale.

So, how is clinical pain measured in horses? There are a number of objective and subjective methods available to assess pain in horses. Objective methods, including physiological data (appetite, heart and respiratory rates, blood pressure and temperature), endocrine changes, nociceptive tests and locomotor activity, are less affected by the observer's judgment. However, they are non-specific and may be intrusive or invasive, costly (if they require equipment) and time-consuming, and the results may not be available immediately (Graubner et al, 2011; Barreto da Rocha et al, 2021).

Physiological parameters correlated poorly with the total visceral pain score (Graubner et al, 2011) and their contribution for somatic pain is also questionable. Physiological and neuroendocrine responses are limited for assessing orthopaedic pain if it is not combined with behavioural indicators (Price et al, 2003).

Unlike humans, who can self-report any pain, horses express unpleasant experiences through their behaviour. Although pain-related behaviour assessment is a subjective method, it is non-invasive, does not require equipment, is free of cost and may be performed remotely by video cameras (Torcivia and McDonnell, 2021). Considering convenience, cost, practicality and efficacy, behaviour is the ‘gold standard’ method to assess clinical pain when it is evaluated by a validated pain scale, because it is feasible in all scenarios, including the field, farms, hospitals and experimental environments.

Behavioural pain scales

The most common equine pain-related behaviours under different conditions are well documented (Bussières et al, 2008; Taffarel et al, 2015; Gleerup and Lindegaard, 2016), and the efforts of different research groups to develop numerous scales to assess pain in horses are summarised in Table 1.

The simplest scales for assessing pain are unidimensional scales, like the simple descriptive scale (1 – no pain, 2 – mild pain, 3 – moderate pain and 4 – intense pain), numerical scale (from 1 – no pain; to 10 – the worst possible pain) and visual analogue scale. The visual analogue scale uses a 10 cm long straight line, where the left extreme is no pain and the right extreme is the worst possible pain; the score score is defined by measuring the distance in millimetres from the left extreme to the point marked by the observer. These scales may be simpler and faster when compared to composite alternatives, but they do not encompass the complexity of pain, have limited reproducibility and are not accurate when used by inexperienced evaluators (Graubner et al, 2011; van Loon and Van Dierendonck, 2018). In horses with visceral pain, the visual analogue scale showed lower inter-observer agreement than composite scales (van Loon and Van Dierendonck, 2018). Composite multidimensional scales embrace the sensorial, cognitive and emotional attributes of pain, have greater sensitivity and indicate the most appropriate pain behaviours to be assessed (Graubner et al, 2011; van Loon and Van Dierendonck, 2018).

A behaviour-based pain scale must have proven reliability and validity before being applied in clinical practice or research. Validity guarantees that the instrument measures what it is designed for, and reliability ensures repeatability and reproducibility. Other important properties of a pain assessment instrument are item interrelation, mutual association and homogeneity, to guarantee a good relationship between items and the contribution of each item. Finally, a cut-off score to provide decision-making for analgesic intervention must be defined (Mokkink et al, 2010; Barreto da Rocha et al, 2021).

Table 1 presents the most important scales according to the validation criteria established by the COnsensusbased Standards for the selection of health Measurement INstruments (COSMIN) (Mokkink et al, 2010) and GRADE (Grading of Recommendations, Assessment, Development and Evaluations) (Guyatt et al, 2008) guidelines, which are well-recognised methods supporting the recommendation of a given instrument considering the quality of evidence and strength.

Assessment of somatic pain

Changes in the locomotor system are the main causes of somatic pain, and lameness is the most common clinical sign in horses with orthopaedic injury (Goodrich, 2008), especially in athletic animals (Wagner, 2010).

Unidimensional scales, like numerical scales, present low reliability when used for assessing lameness (Hewetson et al, 2006). However, the repeatability and reproducibility of an overall score (given by comparing the method −1 – worse, 0 – equal, 1 – better and 2 – improvement in all movements) showed better results than the classical lameness score from 0–10 in horses with osteoarthritis (Fuller et al, 2006).

Back pain is a common cause of somatic pain, especially in competition horses (Fonseca et al, 2006). Objective assessments like electromyography (Lesimple et al, 2012), kinematics (Wennerstrand et al, 2004), thermography and ultrasonography (Fonseca et al, 2006) can facilitate recognition and quantification of pain and enable an efficient diagnosis of thoracolumbar spinal injuries. However, these methods can have a high cost, are impractical and are not applicable in the field.

A protocol has been proposed to grade back pain based on the pain response, muscular hypertonicity, thoracolumbar joint stiffness and overall physical dysfunction of the horse, but it still requires validation (Mayaki et al, 2020). Multidimensional pain scales are more sensitive for assessing pain of variable intensity and chronic pain, which is difficult to diagnose, including head and/or dental pain.

The instruments that are closest to meeting the COSMIN (Mokkink et al, 2010) and GRADE (Guyatt et al, 2008) guidelines are the Horse Grimace Scale (Dalla Costa et al, 2014), the Composite Pain Scale (Bussières et al, 2008), and the Unesp-Botucatu Horse Acute Pain Scale (Taffarel et al, 2015; Barreto da Rocha et al, 2021) (Table 1). The Horse Grimace Scale was initially developed in horses undergoing orchiectomy (Dalla Costa et al, 2014) and was later used for ovariectomy (Rowland et al, 2018), laminitis and thoracolumbar spine injury (Guedes et al, 2024), and ophthalmic (Jodzio et al, 2024) and dental disorders (Sidwell et al, 2025). The Horse Grimace Scale failed to identify pain in horses experiencing equine gastric ulcer syndrome (Ferlini Agne et al, 2023); therefore, other signs like increased girth sensitivity (Millares-Ramirez and Le Jeune, 2019), pawing and aggressiveness before alimentation (Malmkvist et al, 2012), self-mutilation (McDonnell, 2008) and decreased athletic performance might be indicative of equine gastric ulcer syndrome.

In a systematic review, the Horse Grimace Scale showed the highest level of evidence compared to all other facial scales developed for horses (Evangelista et al, 2022); however, a cut-off point indicative of providing analgesia has not yet been calculated. In addition, the association between the Horse Grimace Scale and the Equinosis Lameness Locator was inconsistent in predicting mild orthopaedic pain through lameness assessment before races (Anderson et al, 2023).

The Composite Pain Scale (Bussières et al, 2008) was shown to be the most accurate and reliable pain scale compared to the Horse Grimace Scale (Dalla Costa et al, 2014), Equine Utrecht University Scale of Facial Assessment of Pain (Van Dierendonck and van Loon, 2016) and Equine Pain Scale (Gleerup and Lindegaard, 2016), for correctly identifying horses submitted to experimentally-induced acute arthritis, according to movement asymmetry, using objective gait analysis in horses. However, Composite Pain Scale scores may underestimate pain, as they did not coincide with movement asymmetry in mild lameness (Ask et al, 2022).

Assessment of visceral pain

The most validated and easiest scale to assess acute abdominal pain syndrome is the Equine Acute Abdominal Pain Scale (Maskato et al, 2020). This is a simple and straight-to-the-point instrument that meets the demand for a quick evaluation in emergency cases. The score to discriminate colic cases vs controls is 0.6, mild vs severe pain is 2.8, surgical treatment is 3.5 and death is 4.5 out of 5, providing relevant information regarding clinical conduct, prognosis and therapeutic strategies; a table demonstrating this was produced by Maskato et al (2020). A more complex and time-consuming instrument that can be used to assess acute colic pain is EQUUS-COMPASS (van Loon and van Dierendonck, 2016).

Pain assessment in foals

There are two reasonably well-validated pain assessment instruments in foals. The Foal Composite Pain Scale (Lanci et al, 2022) is currently the most widely validated full-body behaviour instrument available to assess pain in neonates (13±12 [1–47] months old). The items were included according to the literature and expertise and resemble previously reported instruments in adults, complemented with the foal's specific behaviours. The analgesic intervention score was included to guide decision-making to provide analgesia (Table 1).

The Equine Utrecht University Scale for Facial Assessment of Pain in Foals assesses acute pain using facial expressions in neonates (<14 days old) and older foals (15–180 days old), with a cut-off point of >3 (van Loon et al, 2020).

Assessment of chronic pain

If acute pain assessment is a challenge, chronic pain assessment is more so. Like in other animal species, musculoskeletal disease is the principal source of chronic pain in horses and, consequently, lameness is the most frequent clinical sign (Jönsson et al, 2013; van Weeren and Back, 2016; Dyson et al, 2018). Musculoskeletal disorders are generally age-dependent and under-treated (van Weeren and Back, 2016).

Behavioural changes elicited by chronic pain are gradual and subtle, which is why the owners or caregivers are important allies for chronic pain assessment in companion animals (Matsubara et al, 2022). This scenario may be different in horses because owner-reported lameness is low (Ireland et al, 2012).

The Equine Brief Pain Inventory for Owner Assessment of Chronic Pain Due to Osteoarthritis may fill this gap. This scale showed face validity and readability, for use in 5 minutes by owners with limited veterinary knowledge. It shows promise as a tool for assessing osteoarthritis-related chronic pain once fully validated (Howard et al, 2024).

To date, the Musculoskeletal Pain Scale has the highest level of validation to assess musculoskeletal pain. This scale can be used by caretakers and veterinarians, and it incorporates facial expression, posture, head height, weight-bearing and shifting (Auer et al, 2024).

Limitations of behaviour-based pain scales and suggested improvements

The literature makes clear the questionable contribution of some real-time physiological variables (appetite, heart and respiratory rates, digestive sounds and temperature) to assess pain in horses (Bussières et al, 2008; Barreto da Rocha et al, 2021), and their importance requires further investigation. It seems that some of these variables, if not all, might be excluded from pain scales without compromising the construct because they are usually intrusive, time-consuming and cumbersome, and are sometimes not measurable according to the horse's behaviour. Their exclusion would facilitate pain assessment without disturbing the horse.

The second decision is to make the scales as short as possible. The Horse Grimace Scale, Unesp-Botucatu Horse Acute Pain Scale and Equine Acute Abdominal Pain Scale for acute pain and Musculoskeletal Pain Scale for chronic pain are already heading in this direction (Table 1). Conversely, the Composite Pain Scale and Equine Utrecht University Scale for Composite Pain Assessment are long and demand more time. The Equine Utrecht University Scale for Composite Pain Assessment items have not been investigated in isolation to check if they are deemed necessary. In fact, some of the Composite Pain Scale items are redundant and could be excluded (Barreto da Rocha et al, 2021).

Both the Composite Pain Scale and Unesp-Botucatu Horse Acute Pain Scale were critically tested to assess pain after orthopaedic and soft tissue surgery. Although they showed intra- and inter-observer reliability and some validity attributes that met COSMIN criteria, their items were poorly associated and their sensitivity was weak, suggesting that they did not achieve the ideal standards of COSMIN. However, they are the only full-body postoperative pain behaviour instruments to which the COSMIN checklist has been applied to date. The takehome message for the Composite Pain Scale and Unesp-Botucatu Horse Acute Pain Scale is that they require further refinement by excluding redundant or unnecessary items.

Based on this premise, the Composite Pain Scale and Unesp-Botucatu Horse Acute Pain Scale behavioural items were merged, mined and weighted in a single instrument using a predictive model (Trindade et al, 2023). The diagnostic capacity of the short model, including 25% of behaviours from both instruments, was excellent (area under the curve 99% vs 85% for the Unesp-Botucatu Horse Acute Pain Scale and 89% for the Composite Pain Scale). This approach was an encouraging method to ameliorate the current instruments.

The Horse Grimace Scale showed a high level of evidence for measurement properties (Evangelista et al, 2022). A classifier used to estimate the pain status based on the Horse Grimace Scale showed that some facial action units (like ear position and orbital tightening) contributed to identifying pain in horses, while mouth strain and strained nostrils did not (Dalla Costa et al, 2018). Therefore, the Horse Grimace Scale may require refinement, like the Composite Pain Scale and Unesp-Botucatu Horse Acute Pain Scale. Studies on the Horse Grimace Scale also used high-quality images (Dalla Costa et al, 2021) that do not express locum live situations. The horse's face should be accessible without manipulation of the animal to avoid disturbances. Facial scales are liable to missing data when some facial action units are not assessable, or the quality of the images is unsatisfactory as a result of restricted visibility. Facial expression can vary quickly and may not represent the real-time evaluation. Facial scales may also not be effective in detecting certain types of pain, because they do not encompass more complex behaviour repertoires related to different anatomical, sensory, motor, emotional and cognitive dimensions in the way that full-body behavioural pain scales do. Another limitation of facial pain scales is that breed influences superior eye wrinkles in horses, and tension above the eye is one of the items included in Horse Grimace Scale scores (Schanz et al, 2019).

Factors influencing and confounding behavioural pain assessment

Multiple confounding factors can interfere with pain assessment, including human presence (Torcivia and McDonell, 2021). Human presence may cause under or overestimation of facial pain expression in lambs (Guesgen et al, 2016), rodents (Sorge et al, 2014) and rabbits (Pinho et al, 2023) that are truly experiencing or not experiencing pain respectively, and in full-body behaviour expression in rabbits (Pinho et al, 2023) and horses (Torcivia and McDonell, 2021). Torcivia and McDonell (2021) claimed that the caretaker visits interrupted disturbance behaviour in horses; however, only the effect of a human presence outside the box and the examination or treatment administration were not evaluated separately. These factors may inhibit disturbance behaviour to a greater extent than only the human presence. Analysis only of the human presence effect without concurrent intrusive assessments is necessary to confirm that disturbance behaviours are inhibited. Until this information is available, whenever possible, pain assessment should ideally be performed remotely using video cameras.

Other confounding factors include the presence of insects, which augmented tail, ear and head movements in donkeys (de Oliveira et al, 2021a). The time of day, anaesthesia and analgesia influence spontaneous pain behaviours like walking, looking out the window, resting the pelvic limb and resting standing still (Trindade et al, 2021), and the residual effect of anaesthesia may lead to overestimation of the Composite Pain Scale and Equine Utrecht University scale for facial assessment of pain scores for 3 hours (Reed et al, 2023). Post-anaesthetic sedation may influence both body behaviour (Trindade et al, 2021) and facial scales (de Oliveira et al, 2021b).

Although previous experience or familiarity with the pain assessment instruments is advisable (van Loon and van Dierendonck, 2019), the use of scales optimises the ability of less experienced observers to measure pain. When assessing a horse experiencing intense pain, students indicated less analgesic intervention according to their opinion (66%) than according to their Unesp-Botucatu Horse Acute Pain Scale scores (96%), and when using the Unesp-Botucatu Horse Acute Pain Scale cut-off scores, they perceived pain demanding analgesic intervention similarly to an expert (Oliveira et al, 2024).

The future

Algorithms for merging, mining and weighting pain behaviours have been applied to the Horse Grimace Scale (Dalla Costa et al, 2018), Composite Pain Scale and Unesp-Botucatu Horse Acute Pain Scale (Trindade et al, 2023). Automatic recognition of pain based on artificial intelligence and machine vision algorithms was developed by training a convolutional neural network (Lencioni et al, 2021). A software algorithm accurately recognised the early signs and severity of experimentally-induced visceral pain using accelerometry to detect the activity index (Eerdekens et al, 2024). Although using artificial intelligence is the next logical step, these technologies have not yet been tested in the field.

Useful sources of information on pain assessment

After developing pain assessment scales, the subsequent step is to popularise and facilitate their use beyond the academic and scientific world. To this end, various resources were developed:

The Vetpain app can be used to assess acute and chronic pain in horses and all laboratory and domestic mammals, as well as sedation in horses and dogs. Using Vetpain, veterinarians, students, nurses, animal caretakers and owners can also assess videos demonstrating examples of all behaviours present in the pain scales to improve their learning, accuracy and reliability. Videos are available for training and teaching, where the users can check their knowledge according to the template before using the scales on animals. These resources are also useful for teaching purposes. Users can evaluate pain in their animals or animals they are caring for with automatic calculation of scores. Data may be shared among the staff, and because the data are stored, previous pain assessments can be compared to current ones to verify the efficacy and duration of analgesic therapy. The indicative cut-off scores offer assertive information to orient the clinician's decision to provide analgesia for animals experiencing pain, minimising the risk of insufficient analgesia.

A further free source for learning and teaching how to assess discomfort behaviours is the Equine Discomfort Ethogram, which comprises a set of 73 illustrative videos of discomfort behaviours of various intensities, anatomical origins and clinical conditions (Torcivia and McDonnell, 2021). A useful free source for training to identify facial expressions is EQUIFACS (Wathan et al, 2015). Ideally, a metrology instrument should be short, and easy and quick to complete; the simpler the instrument is, the easier it will be to use and the greater the chance of it being used in practice.

Conclusions

This article provides a comprehensive review of the available pain scales to allow readers to discern which scales are appropriate for each type of pain and choose the best option based on simplicity and validity robustness. Accurate pain recognition guides decision-making towards analgesia. The selection of the most appropriate full-body and facial pain assessment scale in horses and foals should be based on the type, progression and location of pain. The free sources available for equine pain assessment facilitate scoring the pain intensity and choosing the best analgesic technique.

KEY POINTS