Successful nutritional control of scratching and clinical signs associated with adverse food reaction: A randomized controlled COSCAD'18 adherent clinical trial in dogs in the United States

Abstract Background Adverse reactions to food are a common dermatological condition in dogs, requiring nutritional intervention using a novel or hydrolysate protein‐based food. Objective To evaluate a therapeutic food containing egg and phytonutrients in dogs with food allergies using an activity monitor and core outcome set for canine atopic dermatitis (COSCAD'18) guidelines and in a controlled double‐masked, multicenter, prospective clinical trial. Animals Adult dogs with a history of adverse food reaction as diagnosed by a food elimination trial were recruited from general practices. Methods After a 21‐day baseline period, dogs were randomized to test or positive control (hydrolyzed protein) food for 21 days. Owner (pruritus visual analog score [PVAS], coat quality, food acceptance, and satisfaction) and veterinarian (canine atopic dermatitis lesion index [CADLI], physical examination) assessments were completed on days 0, 21, and 42. Dogs wore a collar‐mounted activity monitor to record scratching and shaking behavior throughout the study. Statistical analysis included within‐group comparison to baseline and between‐group comparison at study end using a significance threshold of alpha = 0.05. Results At the end of the treatment period, all results were similar between groups for CADLI, PVAS, owner satisfaction, activity, and questionnaire data. Scores for hair dullness, brittleness, amount of dandruff, feces quality, and food acceptance were positive and not statistically different between groups. Conclusions and Clinical Importance The therapeutic test food was well‐accepted and efficacious in managing signs of adverse reactions to food compared to baseline as well as compared to the positive control food.

dullness, brittleness, amount of dandruff, feces quality, and food acceptance were positive and not statistically different between groups.
Conclusions and Clinical Importance: The therapeutic test food was well-accepted and efficacious in managing signs of adverse reactions to food compared to baseline as well as compared to the positive control food.
activity monitor, canine adverse food reactions, egg, novel protein, pruritus

| INTRODUCTION
Canine adverse food reaction (CAFR) accounts for about 10% of dogs presenting for pruritus or allergic skin disease 1 and 15% of those may be allergic to chicken whereas only 4% seem to be allergic to egg. 2 Clinical signs of CAFR typically are dermatological, gastrointestinal (GI), or both and affected dogs can be successfully managed using either alternate or hydrolyzed protein sources. 3,4 Controlled studies of CAFR are uncommonly reported in the literature and are not standardized. Lack of consistency in the design of interventional dermatology studies in dogs led to the proposal of the core outcome set for canine atopic dermatitis (COSCAD'18) which includes 3 endpoints: a veterinary assessment of skin lesions such as the canine atopic dermatitis lesion index (CADLI) or canine atopic dermatitis extent and severity index (CADESI), an owner-assessment of scratching behavior such as the pruritus visual analog score (PVAS), and an owner global assessment of treatment efficacy (OGATE). 5 We used COSCAD '18 criteria to evaluate 2 foods because the criteria could readily be applied to dogs with CAFR as well as dogs with atopic dermatitis.
The major advantage of COSCAD'18 is that it allows for comparison of results between studies by standardizing outcomes.
A drawback is that the measures used are subject to biased reporting by the veterinarian or a pet owner or both. Therefore, we included a wearable device to provide objective data. We previously described behavior algorithms for dogs based on continuous, high-resolution accelerometer data collected by collar-worn activity monitors. 6 The algorithms are sensitive, specific, and >99% accurate in quantifying scratching and shaking in dogs based on validation against >500 video assessments of scratching and shaking in a population of >300 dogs. 6 Activity monitors are capable of continuously and objectively quantifying behaviors in dogs 24 hours per day, 7 days per week, and have been used successfully in a nutrition intervention study. 7 Our objective was to conduct a randomized controlled clinical trial using COSCAD'18 criteria and activity monitor data to evaluate the efficacy of a therapeutic food containing egg and sources of phytonutrients compared to a hydrolyzed protein control food in managing clinical and owner-assessed pruritic signs in dogs with CAFR in a general practice setting.

| Ethics
This study was approved by the sponsor's Institutional Animal Care and Use Committee (CP824a, 11/17/2018) and was performed in accordance with the sponsor's animal welfare policy. Written, informed owner consent was obtained before the commencement of any study activities.

| Patients
Recruitment occurred between September 2019 and March 2020 and included dogs ≥1 year of age if they were free of concurrent disease, had a history of GI, signs (tenesmus, diarrhea, or soft feces) with or without dermatological signs (erythema, scratching) related to an adverse reaction to food diagnosed by resolution of signs after a feeding trial and were currently stable for these clinical signs as determined by the attending veterinarian; patients with known allergy to egg were excluded. Recruitment focused on dogs with cutaneous adverse reaction to food but those with concurrent GI manifestations were permitted provided the patient's GI signs were not consistent with inflammatory bowel disease (eg, undetermined weight loss, hypoproteinemia). Sample size calculations were made assuming a 40% change from baseline for the response variables. Monoclonal antibody-based medications and intermittently-dosed immunosuppressive agents were not allowed; other prior medications were allowed if owners were willing to keep the treatment regimen constant throughout the study. Pet owners were recruited by veterinarians from 11 general practices throughout the continental United States; health was assured based on complete physical examination, CBC, serum biochemistry, serum total T4 concentration, and heartworm test at study entry and a CBC and serum biochemistry profile at study completion. No socialization or enrichment activity restrictions were placed on the dogs.

| Study foods
Details of the macronutrient content and ingredient list of the study foods are provided as supplemental information (Tables S1 and S2); Patients were randomly assigned to either the TTF or PCF in a 1 : 1 allocation ratio using an electronic data capture system (Vision V9P3, Prelude Dynamics, Austin, Texas). Study foods were provided in color-coded bags to mask food identity to all study participants. Dogs included in the analysis entered the study on a variety of foods, but all foods were novel or hydrolyzed protein foods or other therapeutic foods. Supplements or treats were not allowed and dog owners were instructed to maintain the dog's feeding routine. The volume of food offered was based on the dog's current weight, caloric content of the study foods, and input from the veterinarian.

| Activity monitor
All dogs enrolled in the study were provided with a collar-mounted activity monitor which used triaxial accelerometer technology to generate a continuous record of the dog's motion throughout the baseline and treatment periods of the study. After collection, the data were processed using validated algorithms to quantify dermatologic related behaviors including scratching and shaking; sleeping and sleep quality (based on sleep disruptions) also were measured. 6 Study days in which a minimum of 20 hours of data from the activity monitor was obtained were considered "qualifying days" and included in the statistical analysis. Behavior data were summarized as the total duration of each behavior in seconds per day whereas sleep quality was assessed on a scale of 0 (frequently interrupted) to 100 (uninterrupted).

| Data collection
Study entry was considered day 0; questionnaire data from day 21 was considered baseline and compared to day 42 (end of study).
Activity data were recorded on a continuous basis, the baseline period spanning days 0-21, and test period days 22-42. At each visit, veterinarians completed the CADLI and other questionnaires. For the CADLI, grades were assigned reflecting the degree of alopecia/ lichenification/hyperpigmentation and erythema/excoriation/erosion for each of 5 body regions on a scale of 0 (none) to 5 (severe) yielding 2 subtotals that were added together for a maximum score of 50. 8 Scores <8/50 were considered clinically normal. 5 At the same time points, owners completed the PVAS using the scale of 0 (none) to 10 (severe). 9,10 Scores <3.6/10 were considered clinically normal. 5 Owners were asked to score overall skin and fecal quality and answered a series of questions regarding their dog's quality of life based on a questionnaire developed and validated for dogs with atopic dermatitis. 11 Quality of life responses consisted of a 5-point Likert scale ranging from strongly agree to strongly disagree. Other questions were based on a score of 1-100. Fecal quality was graded 1 (watery) to 5 (firm) using a scale that included both pictures and word descriptions. A successful food transition was defined as a transition to the study food without experiencing an adverse event or reporting poor palatability within 7 days of starting the study food, as reported by either the veterinarian or the owner.

| Statistical methods
All data were analyzed by a biostatistician (J. Brejda) who was blinded to the identity of the test and control foods and using Statistical Analysis Software (SAS, version 9.4). Differences in mean animal age and weight among dogs assigned to PCF and TTF were analyzed using a 2-sample t test. Distribution of body fat index (BFI) and sex between dogs enrolled into the PCF and TTF groups was analyzed using Fisher's exact test. Questionnaires that provided a numerical ranging from 0 to 100 were assumed to be continuous and normally distributed and were analyzed using a linear mixed model with food, day, and food Â day as fixed effects in the model. The correlation between the repeated measures (day) was modeled by fitting an appropriate covariance structure selected using the corrected Akaike information criterion (AICC) and Bayesian information criterion (BIC) fit statistics.
A simulation-based adjustment was used to control for inflation of the Type I error rate resulting from multiple comparisons. Trends over days were tested using orthogonal polynomial contrasts for linear and quadratic trends. The pet quality of life questionnaire had nominal ordinal responses such as strongly agree, agree, neutral, disagree, or strongly disagree and was analyzed using the Cochran-Mantel-Haenszel test with modified ridit scores. Separate analyses were performed for each day the questionnaire was completed.
Values from wearable data for each activity during the 3-week baseline period were averaged to produce a single mean value for each animal. Values from the treatment period were averaged at weekly intervals for each animal. The mean baseline values were subtracted from weekly treatment period values for each activity to calculate a period change from baseline (PCFB) value for each animal and week. The PCFB values were analyzed using a general linear mixedmodel with food, week, and the food Â week interaction as fixed effects. To account for the correlation between repeated measurements made at weekly intervals, 6 common covariance structures were fit to the data, and the AICC fit statistic was used to select the best covariance structure for each activity. To control for multiplicity, a simulation-based adjustment (ADJUST = SIMULATE) was used to control for inflation of the Type I error rate. To test for trends over time, orthogonal polynomial contrasts for linear, quadratic, and cubic effects were performed. Differences were considered significant when P ≤ .05.

| Dogs
Patient signalment and characteristics are summarized and provided as supplemental information (Table S3). No significant differences were found between groups with respect to sex, weight, age, or breed. Forty-four dogs were screened, 35 were randomized and received study foods, and 32 dogs completed the study and were    Table S4.

| Canine atopic dermatitis lesion index
No significant differences in CADLI scores were found between and within the PCF and TTF groups at end of baseline and end of study, and results are shown in Figure 1A. Most dogs entered the study with a CADLI <8 (24/32; 75%), indicating a good level of control. 5

| Owner assessments
Outcomes are reported in Tables 1 and 2

| Activity data
Dogs tolerated the use of the activity sensors and owners did not report any issues with their pets wearing the device. Activity data could not be obtained from 4 dogs (2 PCF, 2 TTF) for technical reasons. Compliance was excellent: the average number of qualifying days (days with >20 hours of wearable data) was 75% for the baseline period and 85% for the treatment period. Comparing change from baseline between treatment groups (PCF vs TTF), no significant differences were found in any of the evaluated behaviors: scratching, shaking, sleep duration, and sleep quality index (Table 3).

| Other metrics
Veterinarians were asked a series of questions to evaluate the dog's overall skin and hair coat health. No significant differences were found between or within groups or among time points (Table 4). Successful transition to the study foods was higher for the TTF (14/16, 88%) than the PCF (15/19, 79%) group but was not significantly different. Owner responses to food-related questions are reported in Table 5. For enthusiasm, owners of dogs assigned to the PCF felt their dog was substantially less enthusiastic toward that food than their previous food (P = .04). All other measures were not different between or within groups.

| DISCUSSION
We documented similar control of signs associated with CAFR between foods in owner-dog pairs managed in general practices in the United States using a core outcome set of measurements developed and endorsed by veterinary dermatologists in conjunction with an activity monitor. Similarly, we chose the CADLI, PVAS, and owner evaluation results as primary outcome measures but also used activity data and veterinary outcomes to support these findings. A shortcoming of our study was that the population of dogs was not described based on the specific source of their food allergy. A practical approach for general practitioners is to recommend avoidance of common food allergens by prescribing foods with novel or hydrolyzed protein or both, and the results of our study support that approach. To our T A B L E 4 Veterinary reported outcomes for skin and coat health willingness to recommend foods to a friend or colleague (>50%).
One limitation of our study is that the elimination trial was performed previously and not as part of the trial. As such, we were not able to set the conditions of the food elimination trial. This was because a prospective trial to recruit newly diagnosed cases of CAFR would take a long period of time. We controlled for this issue by recruiting animals where the veterinarian, and not the owner, attested to the presence of the food allergy and history of a food elimination trial. A second potential limitation is the possibility of Type II error. To minimize this possibility, sample size calculations were made assuming a 40% change from baseline for the response variables. This is the magnitude of the response anticipated as a result of feeding the therapeutic food. The anticipated response was observed over time, which from a clinical standpoint was the desired outcome because it showed that the therapeutic food was beneficial. However, it is possible that the magnitude of the difference between the 2 foods was not as large as the magnitude of change over time. As a result, the sample size needed to show a significant change over time was insufficient to demonstrate significant differences between the 2 foods, resulting in a Type II error.

| CONCLUSION
In summary, we successfully used COSCAD'18 criteria and activity monitors to provide evidence for similar management of clinical signs associated with CAFR by 2 foods: a commercially available hydrolyzed proteinbased food and a therapeutic food with egg, rice, and sources of polyphenols.

ACKNOWLEDGMENT
Funding was provided by Hill's Pet Nutrition.