Behavioral phenotyping of cancer pain in domesticated cats with naturally occurring squamous cell carcinoma of the tongue: initial validation studies provide evidence for regional and widespread algoplasticity

Introduction

Pain is a common symptom of human head and neck cancer (HNC). Approximately three in five HNC patients report pain (Macfarlane et al., 2012). Effective pain relief is not achieved with currently available analgesics (Portenoy, 2011; Ripamonti et al., 2012), and the commonly used opioids are addictive, cause significant side effects, and have abuse potential (McNicol et al., 2003). The need for improved HNC analgesics is underscored by the fact that each year, HNC is diagnosed in more than 650,000 people worldwide (Bray et al., 2018). Unfortunately, translational pain research based solely on induced rodent models and then human clinical trials is not efficiently producing new analgesic therapies (Rice et al., 2008; Mogil, 2009). Using an animal model in which the disease is naturally occurring and similar to the human condition may be more predictive of human clinical efficacy (Lascelles & Flecknell, 2010; Klinck et al., 2017; Lascelles et al., 2019). Feline oral SCC (FOSCC) is one of the most common cancers in cats. While FOSCC is not reported to be virally-induced, its biological activity is quite similar to human papilloma virus-negative head and neck SCC (HPV negative HNSCC) in that both are locally aggressive cancers for which: (1) therapeutic resistance is common, and (2) local tumor progression tends to be a common cause for early disease-specific morbidity and mortality (Bilgic et al., 2015).

While the discomfort caused by FOSCC has not been well-studied, veterinary oncologists anecdotally and widely recognize this disease as being a source of considerable pain, which appears to be clinically similar to that pain which is experienced by people with HNC. Thus, FOSCC could be a useful translational, naturally occurring model of oral cancer pain. However, to be a useful model, robust outcome measures are needed, and currently, none have been developed or described. To address that gap, the goal of this study was to develop and assess methods to measure FOSCC-associated pain. To ensure the greatest translational value, it is important that the outcome measures developed for FOSCC-associated pain closely mimic those used in human medicine. We therefore evaluated three different types of pain and/or sensitivity assessment methods: client questionnaires, clinician-based pain scales and quantitative sensory testing (QST).

Several methods and instruments have been used in human HNC patients for measuring pain, including QOL questionnaires (Niv & Kreitler, 2001; Infante-Cossio et al., 2009). In veterinary patients, pet owner-reported questionnaires have been used to measure QOL and pain-associated behaviors in canine cancer patients (Yazbek & Fantoni, 2005; Lynch et al., 2011; Giuffrida, Farrar & Brown, 2017; Reid, Nolan & Scott, 2018). However, none of these tools were developed specifically for orofacial cancers, or cats. In this study, an owner-reported QOL questionnaire, called the Canine Owner-Reported QOL Questionnaire (CORQ), originally designed to assess QOL in dogs with cancer was used as a guide to construct a tool for measuring QOL in relation to FOSCC, and termed the FORQ—Feline Oral Cancer Owner-Reported QOL Questionnaire (Giuffrida et al., 2018); for ease of reading, it is hereafter referred to (in-text) as the FORQ/CLIENT, to provide readers with a quick reminder that this is the client (pet owner) questionnaire.

Attending physician assessments have been used in human pain medicine to complement patient and caregiver reports; a similar approach can be used in cats, with assessments performed by attending veterinarians (Hadjistavropoulos et al., 1997; Zwakhalen et al., 2004; Kunz et al., 2007). Clinician-completed pain scales have been developed for acute pain in cats (Brondani, Luna & Padovani, 2011; Brondani et al., 2013; Calvo et al., 2014; Reid et al., 2017); instruments have also been developed for evaluating chronic osteoarthritis pain in cats (Lascelles et al., 2007; Benito et al., 2013a; Benito et al., 2013b; Gruen et al., 2015; Klinck et al., 2018b; Klinck et al., 2018a) but none have been developed for the assessment of long-standing orofacial cancer pain. Recently, Stathopoulou et al. (2018) used a modified multifactorial composite pain scale (MCPS) from the Universidade Estadual Paulista Campus de Botucatu (UNESP-Botucatu) to identify the analgesic effect of buprenorphine in cats with gingivostomatitis. Based on their positive results, we felt there was a reasonable chance that a similarly modified scale might also provide a robust measure of feline oral cancer pain; thus, additional modifications to the modified UNESP-Botucatu MCPS scale were applied and studied here; in this manuscript, it is referred to hereafter as the UFEPS/VET to reflect the origin of the scale (see literature citation, where the UNESP-Botucatu MCPS is referred to as the UFEPS) and to quickly orient readers to the fact that this is the questionnaire being administered by the veterinary team (vs. the FORQ/CLIENT, which is the pet owner administered tool) (Belli et al., 2021).

In addition to use of clinical metrology instruments such as the UFEPS, complementary information can be gathered through QST. This is potentially useful because pain affects multiple dimensions, and not necessarily all equally (Lascelles et al., 2019). Mechanical and thermal QST represent methodologies that have been used to measure local and distant sensitivity that occurs in humans, and in rodent models (Matos et al., 2011; Zaslansky & Yarnitsky, 1998; Ye et al., 2017). Such methodologies have also been used in dogs with cancer (Monteiro et al., 2018). Electronic von Frey (EVF) and Cochet-Bonnet (COBO) aesthesiometers are examples of mechanical QST (mQST) instruments that have been used for testing cutaneous and/or joint pressure sensitivity, and corneal sensitivity in both dogs and cats (Chan-Ling, 1989; Barrett, Scagliotti & Merideth, 1991; Blocker & Van Der Woerdt, 2001; Briley et al., 2014; Machin, Kato & Adami, 2019). To date, these methods have not been used to assess pain/sensitivity in cats with oral cancer.

A combination of all of these tools formed the basis for the present pilot study, which was performed to begin: (1) validating tools for measuring FOSCC associated pain, and (2) defining various features of oral cancer pain in cats. We hypothesized that it is possible to discriminate between states of normal orofacial sensory processing in healthy cats and abnormal sensory processing in cats with oral cancer by using an owner (client)-reported QOL questionnaire (FORQ/CLIENT), clinician pain scoring instrument (UFEPS/VET), and two forms of mQST.

Materials & Methods

Results

Experiment 1: Test –retest reliability

There were 12 healthy cats and five cancer cats enrolled in Experiment 1 (cancer cat #6 was unable to contribute retest data due to logistical challenges with data collection, not due to subject noncompliance). The test –retest reliability was good for each pain assessment instrument (UFEPS/VET, EVF, and COBO). The UFEPS/VET used in the study can be found as Table S2. The ICC was 0.919 (95% CI [0.776–0.971]). A Bland-Altman plot (Fig. 1A) shows the difference between test and retest on the Y-axis and the mean of two tests on the X-axis; bias between test and retest of UFEPS/VET was 0.5556, which indicated a slightly higher score on the first test. Data from the two tests were positively correlated (r² = 0.77, P < 0.0001; Fig. 1B).

For pressure algometry, the ICC for the maxillary measurement site was 0.73 (95% CI [0.24–0.91]) for average measures. At the metacarpal measurement site, the ICC was 0.68 (95% CI [0.10–0.88]). At the intermandibular sites, these values were 0.73 (95% CI [0.40–0.90]) for the left side, and 0.72 (95% CI [0.37–0.89]) for the right side. The Bland-Altman plot showed most of the differences between the two tests lay between the 95% limits of agreement (LoA) (Fig. S2, left panel). The biases were 5.5, 27.9, 35.3, and 13.6 g in the EVF tests on the right and left intermandibular space, right maxilla, and right metacarpus, respectively. The variability was greater at higher thresholds. The correlations between test and retest were statistically significant at all four testing sites (P = 0.0006 –0.0365, Fig. S2, right panel), which indicated good test –retest reliability for EVF.

The ICC for COBO measurements was 0.21 (95% CI [−0.29–0.62]) for single measures and 0.34 (95% CI [−0.81–0.76]) for average measures. Bland-Altman plots of COBO measurements showed the bias between two tests was 0 cm and the difference between two tests were within LoA (i.e., between −1.9 and +1.9) (Fig. S3). The correlation between the two tests was not statistically significant (P = 0.1823), but the CTTs in cats with sublingual cancers were generally higher than control cats.

Since our test –retest experiments have shown good reliability, we used the first measurement for statistical analysis in Experiment 2: discriminatory validity (except for FORQ/CLIENT, which was answered once by the cat owners in the study).

Experiment 2: Discriminatory validity

Preliminary feline oral cancer owner-reported quality of life questionnaire (preFORQ/CLIENT)

There were 16 healthy cats and six cancer cats from whom data were included in Experiment 2. The discriminatory analysis is summarized in Table 2. The mean ± SD of preFORQ/CLIENT total score in cancer cats was significantly higher (49.67 ± 7.6) than healthy cats (3 ± 6.3, P < 0.0001, Fig. 2A). The VAS was significantly lower in cats with sublingual SCC than healthy controls (P < 0.0001, Fig. 2B); remember: the lower VAS suggests lower QOL, which is opposite of how the FORQ/CLIENT scores change relative to QOL. The mean ± SD of VAS in cancer cats was 55.5 ± 23.65 mm, compared to 93.3 ± 7.9 mm in healthy cats. At the cut-off value of <74.5 mm, the sensitivity and specificity of VAS were both 100%, with the YI of 1. As would be expected, the total scores for preFORQ/CLIENT and VAS from all respondents were inversely proportional (r =  −0.77, r² = 0.6, P <  0.0001, Fig. 2C). The mean ± SD of subtotal scores in terms of frequency and severity were both significantly higher in cancer patients (25.83 ± 6.37 and 21.5 ± 4.04, P < 0.0001, respectively, Figs. 2D and 2E) than healthy cats (mean = 1.68 ± 3.52 and 1.25 ± 2.7, respectively). Apart from the category of “interaction”, the total scores in all other categories (i.e., general behavior, activity, and orofacial discomfort) were significantly higher in cancer cats than in healthy controls (P < 0.0001, respectively, Table S4). The Cronbach’s α value of preFORQ/CLIENT was 0.91, which suggests strong reliability (Cronbach, 1951).

Table 2:

Discriminatory analysis of outcome measures.

Outcome measures	Intraclass correlation coefficient	Sensitivity (%)	Specificity (%)	Youden index	Area under curve (AUC) ± SEM	Effect size	Post-hoc power
FORQ/CLIENT	Not available	100	100	1	1 ± 0	6.69	1
VAS	Not available	100	100	1	1 ± 0	2.14	1
UFEPS/VET	0.92	100	94	0.94	0.97 ± 0.04	2.44	1
EVF-right intermandibular space	0.72	100	94	0.94	0.97 ± 0.04	2.8	1
EVF-left intermandibular space	0.73	100	94	0.94	0.99 ± 0.02	2.59	1
EVF-right maxilla	0.73	100	56	0.56	0.8 ± 0.10	1.26	0.81
EVF- right metacarpus	0.68	83	94	0.77	0.95 ± 0.05	2.5	1
COBO aesthesiometer- left cornea	0.21	100	75	0.75	0.92 ± 0.06	2.08	1

DOI: 10.7717/peerj.11984/table-2

Each question item was assigned a number for further analysis (Table S1). By reviewing the heat plot, the most frequent observations in cancer cats were excessive drooling and trouble eating (QF-19 and -20, Fig. S4A); in regards to severity, the mean score in drooling were the highest in cancer cats when compared to other questions (QS-19, Fig. S4B). Analysis of individual questions showed that 11 of 24 questions were able to discriminate between cancer and healthy cats (Table S5 ). Using a cut-off point of <27.5 for the sum of frequency and severity scores, both the sensitivity and specificity of preFORQ/CLIENT were 100% and the YI was 1.

To refine our questionnaire, and define a FORQ/CLIENT that can be used in future research, we excluded those questions with Cohen’s d value < 0.2 or r <  0.3 in Spearman rank correlation coefficient for both frequency and severity. The refined, proposed FORQ/CLIENT is a 19-item questionnaire (Table 1). The removed questions were: had trouble positioning to defecate/urinate (Q-09), lost balance (Q-11), unwilling to be accompanied (Q-16), did not like to be pet (Q-18), and defensive behavior when touching the head (Q-24). The refined FORQ/CLIENT has an improved Cronbach α value (0.92) and the sensitivity and specificity remained 100% when using a cut-off point of <25.5.

Feline orofacial pain scale for cancers (UFEPS/VET)

Compared to healthy cats, cancer cats had significantly higher total scores in UFEPS/VET, indicating that cats with sublingual SCC exhibit more pain-related behaviors (Fig. 3A). Cancer cats had significantly higher scores in miscellaneous behavior (P < 0.0001), mouth palpation (P = 0.0043), comfort (P = 0.0021), activity (P = 0.0349), attitude (P = 0.0075), and appetite (P = 0.0002), as shown in (Fig. 3B). However, head palpation, vocalization, and posture did not show significant differences between healthy cats and cats with sublingual SCC. Cronbach’s α was 0.84, indicating high reliability. With a cut-off point of >3.5 for total score, the YI was 0.94; at this cut-off, sensitivity and the specificity were 100 and 94%, respectively. The AUC was 0.97 ± 0.04, indicating that UFEPS/VET has high discriminatory capacity. The total UFEPS/VET score was positively correlated with the preFORQ/CLIENT score (r² = 0.58, P = 0.0002, Fig. 3C) and the proposed, refined FORQ/CLIENT score (r² = 0.72, P <0.0001).

In the miscellaneous behavioral category, the majority of observations in cancer cats were: ‘laying down and quiet’ (n = 5/6), and ‘licks, ptyalism, or chattering’ (n = 4/6) (see Table S2). In attitude evaluation, five out of six cats with sublingual SCC were uninterested in interacting with the observer and one cat showed both disinterest in the observer and indifference to its surroundings. Cats with sublingual SCC tended to show interest in food, but did not continue consumption after a few bites.

Cohen’s d values for all items in UFEPS/VET were large (>0.8), except for “vocalization” (d = −0.04). Spearman rank correlation coefficient test showed “head palpation” had low correlation to the total score (r = 0.25). After removing “vocalization” and “head palpation”, the Cronbach’s α = 0.88; at a cut-off point of >3.5, the sensitivity, specificity, and YI remained 100%, 94%, and 0.94, respectively, but the AUC increased to 0.98 ± 0.03.

Electronic von Frey and Cochet-Bonnet Aesthesiometer

Cats with sublingual SCC were significantly more sensitive to pressure (i.e., had lower mQST thresholds) at both sides of the intermandibular space (Figs. 4A and 4B, P < 0.0001 on both right and left side), and at the right maxillary region (Fig. 4C, P = 0.0273). The mean differences of the mechanical sensory threshold between healthy and cancer cats on the right intermandibular space, left intermandibular space, and right maxillary region were 158.5 ± 31.02, 152.2 ± 31.45, and 101.1 ± 42.58 grams, respectively. Based on the ROC curve, the sensitivity can reach 100% at either side of intermandibular space, and the specificity is 94%. At right maxillary region, the sensitivity and specificity were 100 and 56%, respectively. Interestingly, decreased thresholds were observed at the metacarpal site (P = 0.0002) in cancer cats compared to healthy controls (227 ± 32.32 versus 378.7 ± 16.43 g, Fig. 4D). The sensitivity and specificity of EVF on metacarpal region were 83.33 and 93.75%, respectively. The AUC ranged from 0.8 to 0.95 at all EVF test sites, which indicates excellent discriminatory ability. In cancer cats, five out of six cats had tumors that originated from the right side and one out of the six cats (FOSCC#4) had a tumor originating from the left side. However, all of the cats had masses that crossed the midline and occupied the base of the tongue. There was no significant difference between the mQST thresholds at ipsilateral and contralateral sides of the intermandibular region in cats with sublingual SCC (Fig. 4E, P = 0.883).

Previous work has demonstrated that healthy domestic short-haired cats have similar CTT values in the right and left eyes (Blocker & Van DerWoerdt, 2001). To reduce procedure time and stress on animals, we only performed the COBO aesthesiometer test on left corneas. In contrast to the threshold value measured by EVF, higher CTT values indicate that a subject is more sensitive to corneal touch. CTT values from the left cornea showed that cats with sublingual SCC were more sensitive than control animals (Fig. 4F, P = 0.0009); on average, healthy cats were able to sustain higher pressures (1.17 ± 0.3 cm shorter filament). At the cut-off point of >2.25 cm, sensitivity and specificity of COBO for discrimination between cancer cats and healthy cats were 100 and 75%, respectively. Based on the ROC curve, the COBO has outstanding discriminatory capacity (AUC 0.92 ± 0.06).

Sex differences have been shown when measuring orofacial sensitivity using QST in humans (Costa et al., 2019). We compared the threshold values measured by EVF and COBO aesthesiometer between healthy castrated males (n = 6) and spayed females (n = 10), but did not identify statistically significant differences between sexes (Fig. S5). There were too few cancer-bearing cats to make a statistically relevant comparison.

Experiment 3: Responsiveness validity

UFEPS/VET, EVF, and COBO measurements were performed before and 30 min after administering an analgesic drug (buprenorphine 0.02 mg/kg, buccally) to six cats with SCC of the tongue. The mechanical sensory threshold at the ipsilateral intermandibular site, and CTT values were significantly improved after buprenorphine administration (Fig. 5, the details can be found in Table S7). The mechanical sensory threshold increased by 66.85 ± 24.48 g (P = 0.04) at the ipsilateral intermandibular site after giving buprenorphine, and after buprenorphine, cancer cats could withstand corneal pressures that were higher (1.17 ± 0.33 cm shorter filament; P = 0.017). Although not statistically significant, other tests (including UFEPS/VET, and EVF at other sites) also showed a trend toward decreased pain/sensitization after providing buprenorphine; this is notable given the small sample size, where lack of statistical significance may reflect true biology or statistical error. Based on post hoc power analysis, with the exception of COBO testing, power was low (< 0.8) for all comparisons in Experiment 3 (Table S7). Nonetheless, results are biologically plausible, and for each assay, the pain score/sensitivity returned to baseline (i.e., pre-buprenorphine levels) by the next morning.

Discussion

This study reports on the adaptation and initial testing of various outcome measures that should enable robust research in a naturally occurring feline model of oral cancer pain. Based on our preliminary studies, the methods described herein appear to have good intra-rater reliability within a single observer; furthermore, we provide evidence that these assays effectively discriminate between healthy and tumor-bearing cats, and some are responsive to analgesic therapy. Interestingly, our results also suggest that cats with tongue cancer not only develop pain in their mouth, but also seem to have regional peripheral, and widespread sensitization; this is consistent with the widespread somatosensory sensitization that is reported in human HNC patients and which can be measured in other animal models of oral cancer pain (Ye et al., 2017).

One goal of this study was to develop methods for assessing pain in a naturally occurring feline model of oral cancer. Pain can affect overall QOL, thus in our study, we attempted to measure QOL rather than health-related QOL, which only focuses on the impact of a specific illness on QOL (Reid, Nolan & Scott, 2018). Given that a canine owner-reported QOL questionnaire for cancer patients had been developed and published, we used this to guide the initial construction of an instrument to evaluate the QOL of cats with oral cancers. Through basing this instrument on the canine cancer QOL instrument, and discussions between the authors, we believed the initial FORQ/CLIENT had sufficient face validity to move forward into testing and refinement. Our initial (preliminary) version of the FORQ/CLIENT showed high reliability and good discriminative function; it distinguished healthy cats from those with sublingual SCC. However, some individual questions’ effect size was small, leading us to omit them from the proposed version of the FORQ/CLIENT. In our study, cats with oral cancers did not have significant mood alteration, reluctance to wake up, or vocalization when resting, and they did have normal interactions with the owners, which is different from observations made in canine cancer patients (Giuffrida et al., 2018). This discrepancy may be due to the natural characteristics of different species. We also found that “less drinking” was not a good indicator in cats with sublingual cancers. Although cats with sublingual cancers decreased their normal food intake, it is known that mouth dryness and pain are common in human patients with advanced cancers. We therefore theorize that increased water intake could possibly indirectly indicate mouth dryness and/or pain in these tongue cancer-bearing cats (Oneschuk, Hanson & Bruera, 2000). Overall, similarly negatively altered QOL has been reported in human HNC patients (Gotay & Moore, 1992; Terrell et al., 2004), which supports the potential translational utility of this outcome measure.

To complement these pet owner derived assessments of QOL, we also tested a clinician-based clinical metrology instrument focused on measuring the impact of oral pain. UFEPS/VET scores were significantly greater for oral cancer-bearing cats than control cats, and UFEPS/VET scores were highly correlated to FORQ/CLIENT responses. The UFEPS/VET assesses various behavioral changes that indicate pain and/or discomfort, not only around the oral region, but also in terms of general comfort, activity and attitude. The Cohen’s d value shows small effect size for vocalization, and the Spearman rank correlation coefficient showed head palpation related poorly to the total score; thus future iterations of the UFEPS/VET might exclude these parameters. Most cancer-bearing cats maintained interest in food; however, tumor-associated dysphagia is known to reduce food intake (which is reflected in the “difficulty eating food” data in FORQ/CLIENT); thus, in future iterations of the UFEPS/VET, we may consider revising the appetite scoring to reflect this disconnect between appetite and ability to successfully prehend food. Additionally, the value of UFEPS/VET may be improved through the incorporation of facial expression assessment; both ear position and muzzle shape have been used to in feline acute pain evaluations (Holden et al., 2014; Reid et al., 2017; Evangelista et al., 2019). Although a clinician-based pain assessment has not been developed specifically for cancer pain in humans, behavioral observation scales produce similar results in nonverbal elderly adults with cancers and chronic pain (Passik et al., 2004; Bjoro & Herr, 2008); this supports the potential translational utility of the UFEPS/VET in cats with oral cancer. Both FORQ/CLIENT and the UFEPS/VET should undergo additional testing including repeating our work, readability testing and additional validity testing, including inter-rater reliability, responsiveness testing with other analgesics, further face validity testing and criterion validity testing.

We used an EVF and a COBO aesthesiometer to evaluate mechanical sensitivity. The EVF thresholds measured on the maxilla of healthy cats were variable; this could be due to the presence of highly sensitive tactile hairs of the vibrissal pad (Gogan et al., 1981; Grant et al., 2013). Even with that variability, we demonstrate here that cats with sublingual cancer were more sensitive to mechanical pressure near the mouth, and, interestingly at a distant anatomic site. Increased sensitivity around the mandibular region, maxilla, and eyes may be partly related to peripheral sensitization driven by locally produced nociceptor sensitizers, but part may also be driven by more central changes affecting multiple branches of the trigeminal nerve (Woolf, 2011; Vardeh & Naranjo, 2017). The three trigeminal branches (i.e., ophthalmic, maxillary, and mandibular nerve) anatomically converge at the trigeminal ganglia, and signaling cross-talk is also possible at the brainstem (Walker, 1990). This pathway of signal transduction explains the maxillary and corneal hypersensitivity we observed. Peripheral sensitization may be due to inflammation, tumor cells infiltrating into the nerve fibers or other tumor-related factors (trauma, acidification, etc.) that increases the responsiveness and activity of nociceptors (Vardeh & Naranjo, 2017). The finding that cats with sublingual SCC had relatively low nociceptive thresholds at their metacarpus (in the absence of a tumor or other injury on their paw) may indicate central sensitization, or a generalized peripheral sensitization (Woolf, 2011; Vardeh & Naranjo, 2017). This is consistent with the finding that mice with experimentally-induced tongue SCC display widespread somatosensory sensitization (Ye et al., 2017).

The QST measures that were studied herein were able to distinguish healthy cats and cats with sublingual cancer. However, it must be remembered that QST results are likely affected by the “pain history”, underlying chronic pain, and stress. To address such potential bias, we performed a subset analysis in age-matched cats that came to the veterinary hospital for regular check-ups and/or vaccination, rather than simply visiting to participate in this study. This analysis supports our conclusion that the observed hypersensitivities were related to oral cancer. To that end, it is important to reflect upon the low ICC values reported herein for both mQST assays. Here, the lack of stability over time is not surprising given the small sample size, and this result is also similar to what has been reported for other forms of QST utilized in companion animals (Knazovicky et al., 2017). Thus, given our existing knowledge, it is important to consider that a well-designed experiment utilizing these assays must include carefully chosen concurrent controls.

To understand how well these QST data recapitulate the human oral cancer pain condition would require comparing our results with QST data from affected people. However, QST has been primarily used in human cancer patients to characterize the somatosensory changes that come with chemotherapy-induced peripheral neuropathy; data on OSCC related pain are lacking (Martland et al., 2020). Similar to the human condition is that pain is often restricted to the local tumor site. Dissimilar is that we measured spontaneous pain, whereas human oral SCC patient seem to report pain most frequently as having been evoked by such activities as eating and speaking.

As an initial evaluation of whether the methods used in the present study are able to detect reduced cancer pain after analgesic treatment, we performed assessments before and after administration of buprenorphine. Buprenorphine is an opioid that has been primarily used in cats for acute surgical pain (Steagall et al., 2009). Oral transmucosal administration has been shown to relieve pain and the maximum plasma concentration occurs about 30 min after application (Robertson et al., 2005; Stathopoulou et al., 2018).While buprenorphine may cause sedation, which could potentially confound our results, we saw significantly reduced sensitivities at both of the ipsilateral intermandibular test sites, and on the corneal surface after buprenorphine administration. This suggests that the instruments described herein are responsive to, and appear able to measure the effectiveness of analgesic interventions. Future validation studies should seek to understand whether the assays are also responsive to other analgesics. For example, it would be logical to test a non-steroidal anti-inflammatory drug; that was not done here due to concern for renal injury, which was heighted by the fact that after these pain measurements were made, each cat underwent general anesthesia for cancer treatment.

There are several limitations to the pilot work presented herein. First, while control cats did undergo thorough physical examination by a trained veterinarian, sedated oral examination and dental radiography were not performed, thus it is uncertain whether these cats were free of any dental disease. Second, while physical restraint for mQST was gentle and minimal, that handling and the environment (a hospital setting) could have affected nocifensive (behavioral) responses. Third, in Experiment 1, the amount of time between measurements (no less than 6-12 h) may have been too short; that interval was chosen because it was viewed as a balance between allowing time between tests while also minimizing the amount of hospitalization required for each cat. That latter point is an important consideration since clinic/hospital visits are potentially stressful to the animal, and that stress could influence test results. Nonetheless, is possible that during the second assessment, the observer may have been influenced by memory of answers to the first assessment. Strategies to reduce such potential bias must be considered for future validation work. Finally, in Experiment 2, our ability to rigorously assess discriminatory validity is compromised by failure to effectively mask the observer to experimental group; therefore, construct validity and responsiveness are subject to bias. Unfortunately, the frequency with which oral cancer afflicted cats display overt clinical signs of disease that are evident even to non-expert observers is high, and thus effective masking of well-trained researchers is viewed by these authors as being an unavoidable limitation of the work. Next steps in validation of these assays will be to larger scale studies, with incorporation of multiple observers such that we can rigorously assess inter-rater reliability.

Conclusions

Together, the methods described in this manuscript can be used to provide what we believe is a comprehensive assessment of orofacial pain phenotypes in cats with oral cancer. If these instruments are to be successfully used for rigorous assessment of cancer-related pain and/or analgesic efficacy, additional large-scale inter-institutional studies should be performed to ensure that reproducible results are achievable. It would also be ideal to understand assay responses in the setting of various cancer treatments (e.g., radiotherapy, chemotherapy, and surgery). Another important finding was that cats appear to have pain at the local/primary tumor site, and also have evidence of widespread somatosensory sensitization. Future work should aim to: (1) reaffirm this finding, (2) more comprehensively characterize the widespread changes and their mechanistic underpinnings, and (3) determine whether similar changes occur in humans, particularly in the absence of distant metastatic disease.

Supplemental Information