Quantitative sensory testing (QST) has been widely employed in the investigation of somatosensory alterations associated with central sensitisation in man (1–3) and evaluated as a means of predicting analgesic response (4). Sensory profiling, permitting sub-grouping of patients, has been demonstrated to identify treatment responders in some pain conditions (5). Quantitative sensory testing in the form of nociceptive response testing is well established as a fundamental outcome measure in pre-clinical analgesic and pain research in non-human species (6–10), and recently QST has begun to be investigated in research units for utility in the assessment of companion animals (11–13), both as a potential veterinary clinical tool and as a pre-clinical measure in spontaneous disease companion animal models. Knazovicky et al. (2016) (14) reported widespread somatosensory gain of function in association with OA in dogs providing evidence of central sensitisation in this population of dogs and strengthening the validity of canine spontaneous OA as a model for human OA. Significant clinically problematic aspects of central sensitisation in man relate to inappropriate perception of normally non-noxious stimuli as painful (15, 16) (allodynia), an aspect of QST which has previously not been explored in dogs suffering from spontaneous OA. Loss of sensory function, identified by increased (e.g., vibration or mechanical) detection threshold, has also been reported in association with central sensitisation in man (15); the assessment of response to the application of non-noxious stimuli to animals may be able to detect such sensory loss.

Environmental factors affecting QST data collected from dogs are not well described. Investigators have reported using a dedicated or quiet space with a relatively constant temperature (17, 18) in order to minimise distraction for the dogs undergoing testing. However, for widespread adoption of the technique as a pre-clinical and clinical tool, QST measurement should be feasible and demonstrate repeatable results in common clinical veterinary settings.

Our aim was to evaluate the utility of a wide range of QST measures in client-owned dogs suffering from spontaneous OA within an environment comparable to a veterinary practice and compare these recordings with data from a matched control group. Using a veterinary practice setting to collect data would be conducive to the ultimate aim of recruiting large numbers of dogs into studies. The studies reported here were part of a larger project (19) aimed at understanding the association between neurophysiological measures of central sensitisation and QST measures. We hypothesised that dogs suffering from OA would demonstrate altered QST measurements compared with normal dogs, showing hyperalgesia and allodynia to noxious and non-noxious stimuli respectively.

In contrast to our hypothesis, OA category was not a major determinant of QST outcome measures for the majority of modalities evaluated and, even when significant differences were detected between dogs with OA and the control group, the separation between the groups was minimal so that at the level of the individual animal the QST modalities were poor diagnostic tests. This was surprising, given that other research groups have identified somatosensory changes indicative of CS in dogs suffering from OA through the application of similar QST methods (12, 14, 17), and that we were able to identify group level alterations in neurophysiological responses, consistent with CS, in the OA and OANSAID cohort studied here (19). One potential factor could be that there is severe data heterogeneity among QST studies in dogs (26). In the few modalities in which OA category was determined to be a significant predictor variable, results were mixed compared with previously reported data. Our results were suggestive of an increased mechanical nociceptive threshold (primary hypoalgesia) at the hip in OA affected dogs compared with controls (in contrast to previous work), however we found a decreased latency to respond to thermal stimulation of the plantar digital pad (secondary hyperalgesia) (consistent with some previous work) (14). One of the novel tests (air puff) demonstrated decreased responses in both OA and OANSAID dogs, and may reflect tactile hypoalgesia. Knazovicky et al. (2016) (14) demonstrated decreased mechanical threshold to both deep pressure and von Frey stimuli at both primary and secondary sites in OA affected dogs, findings which correlate with reported increased mechanical sensitivity in human patients suffering from OA (27, 28).

Thermal hypoesthesia, as evaluated by the CTES, has been reported in dogs suffering from OA, compared with control dogs (12), whilst our results are suggestive of increased heat pain sensitivity using this apparatus. Williams et al. postulated that their results may be indicative of dogs with painful joints not wanting to shift weight and lift the hindlimb in response to the stimulus. One study in humans reported that heat pain thresholds at both primary and secondary sites were not different in OA compared with controls (2), whilst other investigators (29) have reported decreased heat pain thresholds associated with OA (i.e., thermal hyperalgesia). In humans it has been suggested that subgroups of QST phenotype, identifiable by cluster analysis (30, 31), exist in both OA affected (32) and healthy (33) populations, therefore comparing outcome measures between OA and control groups might not be expected to reveal differences between the populations, as inter-individual variability may be relatively more important than disease status. The importance of individual (dog) above other factors as a source of variability in determining mechanical nociceptive threshold in healthy dogs has been reported (34) and may reflect the presence of individual QST phenotypes in dogs.

Knazovicky et al. (2016) (14) determined that feasibility was a significant source of variability of the thresholds recorded for von Frey and heat stimulation at the primary site, therefore it is possible that the generally lower feasibility of testing which we recorded has impacted the measured thresholds, and confounded differences between groups. The proportion of QST assessment attempts in client owned dogs which have been completed with acceptable feasibility has previously been reported in the order of 80% (12, 35), however, only 56% of the first QST sessions in the present study were considered to have acceptable feasibility, although this increased to 73% for second QST sessions. Harris et al. (2015) (34) reported a number of factors which influenced the response rate to a mechanical stimulus, finding that a tip diameter of 2 mm (as used in the present study), and performance of the test with the dog in a sitting position improved the ability to record a response, however, application at both metatarsus and hip sites was not evaluated in that study, and it would have been difficult to achieve application of stimuli to the dorsal surface of the metatarsus whilst the hock was flexed, in a sitting position. Briley et al. (2014) (35) attained a proportion of 83% feasible mechanical nociceptive tests with dogs lying in lateral recumbency, which suggests that aspects of testing in our protocol (which was designed to mimic conditions in a quiet room within a shared building as would be a common scenario in clinical veterinary practice) adversely affect feasibility. The occurrence of stimuli within the shared building, even if not impinging on the testing room, appeared to result in distraction of the dogs from the testing. A large proportion would stand up and approach the door of the testing room in response to hearing entrants to the building. Stimuli related to housekeeping, occurring within the building but outside of the testing room, would cause a noticeable alteration in attention of some of the dogs. Rarely did people walking past the window seem to capture the attention of dogs. It is likely that minimising extraneous stimuli is necessary to achieve good feasibility during testing protocols that are dependent on behavioural outcomes.

Feasibility was only assessed in a minority of dogs within the study, however, the distribution between groups, and demographics of the assessed dogs, were representative of the population as a whole. It is likely that there was insufficient statistical power provided by this number of samples to evaluate an improvement in feasibility from the first to second QST session. However, a change from 56% to 73% acceptable feasibility would suggest a clinically relevant difference and may suggest that familiarity with the test environment is required over a greater period than that offered by the initial 15-min acclimatisation.

Before QST can be recommended as a clinical or translational research tool a thorough quantification of factors affecting the feasibility is necessary. Video analysis of testing protocols, coupled with time locked information on external stimuli, may provide a beneficial overview of factors which need to be studied more closely.

Both the von Frey filaments and the Electronic von Frey Anesthesiometer may be used for this purpose. Study in cats showed that performing QST with both the Electronic Von Frey and Von Frey Filaments is feasible, and that the sensory thresholds measured at the lip and at the stifle with these two algometers are comparable, indicating fair interdevice agreement (36). Due to its practicality, broader numerical range, and reduced need for repeated applications, the Electronic von Frey may be considered a superior option for QST compared to Von Frey Filaments.

Self-report of sensations associated with QST in verbal humans enables the relatively straightforward detection of subtle changes, for example with regard to detection threshold, and permits the evaluation of differing degrees of pain perception, for example pain threshold and pain tolerance. In contrast, behavioural measures in animals must be employed as surrogate markers for awareness of stimulation, and are unable to definitively differentiate between nociceptive responses/reflexes, and the conscious experience of the feeling of pain. Moreover, responses in awake animals may reliably indicate nociceptive (or pain) thresholds, or may be over-ridden by tolerance of the nociceptive stimulus (14) that could vary between dogs. The companion role of dogs and associated desirable trait of calmness/compliance (37) (which may manifest as a willingness to appease humans), may also have ramifications with regard to their response to a nociceptive stimulus, particularly one associated with a human investigator as in the case of a hand-held device.

Decreasing response to the brush with increasing stimulus number may represent learned tolerance to a non-nociceptive stimulus, whilst decreased latency to the CTES with increasing stimulus number may represent a learned response to a nociceptive stimulus. Decreases in threshold assumed to be learned/anticipatory behaviour in association with repeated mechanical nociceptive stimulation have previously been documented (38), therefore it may be preferable to utilise tests employing single stimuli to quantify responses, without the confounding effects of learning.

Body weight has previously been reported to be a significant factor in the determination of mechanical nociceptive stimuli (34, 35), and our results suggest that, additionally, vibration and thermal thresholds also vary with body weight. This finding corroborates the evidence produced in our electrophysiological studies (19), and suggests that publication of normal ranges for QST measures will require a correction factor for body weight.

In summary, the techniques as described in this study suggest that, in a clinical setting, variability in the results between dogs may be substantial enough to obscure any changes associated with CS. Consequently, the methods cannot currently be recommended for evaluating CS in dogs. Other methodology and objective measures such as the nociceptive withdrawal reflex, should be explored. Additionally, the application of stimuli that appear remote from the investigator, as seen with CTES, may produce more meaningful data (in terms of indicated altered central processing of noxious stimuli) compared to those delivered via hand-held devices.