**The Impacts of a Reading-to-Dog Programme on Attending and Reading of Nine Children with Autism Spectrum Disorders**

### **Stefania Uccheddu 1,\*, Mariangela Albertini 2, Ludovica Pierantoni 3, Sara Fantino <sup>3</sup> and Federica Pirrone <sup>2</sup>**


Received: 29 May 2019; Accepted: 23 July 2019; Published: 26 July 2019

**Simple Summary:** The purpose of this research was to compare reading motivation and attitude, as well as reading and cognitive skills, of school-age children diagnosed with Autism Spectrum Disorders (ASD) who attended a 10 session reading programme with and without the presence of a dog. Children who read to a dog had 100% attendance at sessions over the course of the programme versus 75% (range 25–100%) of children attending reading sessions without a dog. In addition, after the programme, they were significantly more motivated and willing to read at home, as perceived by their parents. However, there were no significant differences in scores on reading and cognitive tests either within each group or between groups. Based on these results, we can conclude that reading to a dog can have positive effects on an ASD child's motivation and attitude toward reading. More research is needed to better understand if it can also have positive effects on children with ASD's overall reading and cognitive abilities.

**Abstract:** Poor knowledge is available on the effectiveness of reading to dogs in educational settings, particularly in children with Autism Spectrum Disorders (ASD). In this study, we test the hypothesis that reading to a dog improves propensity towards books and motivation to read after the end of the programme, as well as reading and cognitive skills in children with ASD. The study is a prospective, randomized controlled trial, consisting of testing and re-testing after a 10 sessions reading programme with and without the presence of a dog. Nine Children with ASD (6–11 years old) were randomly assigned to a control (CG, reading without a dog, n. 4) or experimental group (EG, reading to a dog, n. 5). Children's attendance at reading sessions was recorded at each session. Parents' perceptions were evaluated at the end of the programme to detect changes in children's attitudes and motivation toward reading. Psychologist-administered validated reading (Cornoldi's MT2 reading test; test of reading comprehension, TORC; metaphonological competence test, MCF) and cognitive tests (Wechsler intelligence scale for children Wisc IV, Vineland) to all children, at baseline and at the end of the reading programme. Compared with CG children, children in the EG group participated more frequently in the reading sessions, and they were reported to be more motivated readers at home after the programme. However, there were no differences on reading and cognitive tests' scores either within each group of children or between groups. Further studies are warranted in order to understand whether and how incorporating dogs into a reading programme is beneficial to Children with ASD at the socio-emotional and cognitive level.

**Keywords:** dogs; children; Autism Spectrum Disorders; cognition; reading-to-dog programme

#### **1. Introduction**

Launched in 1999 by Intermountain Therapy Animals, the Reading Education Assistance Dogs (READ) is the first programme and still one of the most comprehensive involving animals to strengthen children reading skills [1]. The Reading Education Assistance Dogs® programme improves children's reading and communication skills by employing a powerful method: reading to a dog (R.E.A.D. webpage)

Recent work focused on children has shown that reading-to-dog programmes might reduce blood pressure and offer a nonjudgmental, safe environment in which to practice reading [2,3]. Dogs seem able to offer children a unique type of emotional support in the education setting because they are fully capable of being active, supportive listeners, but are also unable to verbally criticize or comment upon a child's reading abilities [4]. In the wider literature [1], measurements of reading skills have included improved scores on test of reading comprehension (TORC), measures of academic progress (MAP), reading rate, and reading ability [1]. According to Pillow-Price et al. [5], all reading scores for children participating in a reading programme improved significantly. Sorin et al. [6] noted improvements in reading, behavior, confidence, self-esteem, and school attendance with special education students who worked on literacy skills with dogs. Changes in reading motivation may reflect a better reading performance [1]. In Guthrie and Cox [7], engaged and motivated children who opened a book more frequently were also highly achieving in reading abilities because cognitive functioning was powerfully facilitated through interest and motivation [8].

The presence of a dog has already been suggested to reduce physiological parameters of stress (decreased blood pressure [9] and cortisol [10]) in children with autism. A child with autism spectrum disorder improved on the Dynamic Indicators of Early Literacy Skills (DIBELS) and Elementary Reading Attitudes Scale (ERAS) after completing a reading-to-dog programme [11]. Based on the final version of DSM-5 [12], autism is currently counted in one general term, Autism Spectrum Disorders (ASD), with three different levels, level 1 (requiring support), level 2 (requiring substantial support), and level 3 (requiring very substantial support). ASD is characterized by delays in the development of multiple basic functioning including socialization and communication and behavioral challenges (such as rituals and repetitive behaviors) [12]. In the clinical setting, anxiety-related concerns are among the most common presenting problems for school-age children and adolescents with ASD [13]. Recently, one study developed an educational setting in which three Children with ASD read social stories in the presence of therapy dogs [14]. The authors aimed to test the hypothesis that the presence of a therapy dog improves the effectiveness of Social Story method, which is used to communicate clear and detailed information to autistic children on a context, skill, achievement, or concept [15]. Although improvements in these children's indicators of social skills were reported (e.g., increased frequency of the initiations of social interactions and decreased level of prompt needed to provide the expected social response), interpretation of these indicators can be difficult, and the quality of the evidence is still unclear, also due to the small sample size used [1].

Here, we decided to apply recent and innovative psychological approaches to detect potential improvements in reading abilities and changes in behavioral and emotional processes in ASD school-age children's reading in the presence of a dog compared with children reading without a dog. Moreover, we compared attendance at sessions of children and parents' perception of their reading motivation and willingness to read. The main hypothesis was that a social environment enriched by the presence of a dog strengthens the effectiveness of a reading programme in enhancing both reading and cognitive abilities in Children with ASD.

#### **2. Materials and Methods**

#### *2.1. Participants*

A total of nine children in the age range of 6 to 11 years (mean 7 ± 0.45 SE), seven boys and two girls, were recruited from the CTR Esperienze ONLUS (Comunicazione Territorio Relazioni) Cagliari, Sardinia, Italy, where the reading sessions took place.

Informed consent was obtained from parents of all children, who were previously advised by the facility staff members of an experimenter's presence for the videotaping procedure. In signing the consent, parents ensured that there was a clear understanding of the information given to them, and also that they agreed with that and with the disclosure of their personal details. Besides favorable opinion from a clinical psychologist or neuropsychiatry holding also a certification as Board Certified Behavior Analyst® (BCBA®), to be eligible for participation in the programme, children were required to: (a) be diagnosed according to the diagnostic tools described in the DSM-5 and in the guidelines elaborated by the Italian Ministry of Health guidelines. The diagnosis was determined by a multidisciplinary equipe composed of a child neuropsychiatrist, a psychologist specialized in child development, and a pedagogist. The DSM-5 diagnosis also includes new guidelines for categorizing autism by level. There are three levels, each reflecting a different level of support each child needs (from level 1: little support, to level 3: higher support); (b) show lack of initiation of appropriate social response in a given social situation during therapy or free-time activities; (c) have some reading prerequisites, such as the ability to open and browse through a book; (d) be willing to interact with dogs, as evaluated in a preintervention screening; (e) possess basic speaking skills, and (f) immunocompetency. Fear of dogs was considered an exclusion criterion. Diagnosis and severity level have been established by the neuropsychiatry according to [12].

#### *2.2. Reading Session*

Children with ASD were randomly divided into two groups according to demographics characteristics and severity levels expressed in the diagnosis: (1) the experimental group (EG, n. 5: four boys and one girl; mean 7.60 ± 2.30 SE) read a book with a dog present, and (2) the control group (CG, n. 4: three boys and one girl; mean 8.25 ± 1.73 SE) read a book without a dog. Details on age, gender, and level of severity indicated in the diagnosis are reported in Table 1. Groups were homogeneous in terms of mean age and diagnosis. Both EG (experimental group) and CG (control group) were involved in 10 weekly group sessions, run over a period of 70 days. Each session was approximately 30 min in length, without pauses, during which children read a book one-on-one, upon request by the psychologist. A book was selected by the psychologist at the beginning of the programme. Both EG and CG children read the same book. A copy of the book was available for each child. The same psychologist was present for all the sessions for both EC and CG groups. The psychologist, before starting the reading session, reminded the children of the rule of the session. The rules were presented, if necessary, more times during the session, only in oral form (Now we are going to read to the dog. Her name is Bella/Lilli. The dog is pleased to listen to our reading, but we need to respect some specific rules: please, do not be loud, do not run, do not touch the dog since this is going to make the dog fearful. We are not going to pass through the benches during the reading session. We cannot touch the dog during the session, but we can talk with her).

Sessions were performed in the afternoon in order to exclude parental factors/obligations that could impact the child's attendance.

As for the experimental group, two dogs (both neutered females, mixed-breed, 2 and 8 years old) participated in the sessions, one at a time, on alternate weeks. Two dogs were chosen by a team composed of two veterinarians expert in behavior and welfare and a psychologist specialized in animal-assisted intervention. Inclusion criteria considered their kindness and cooperation when handled by children, their interest in people, and absence of any signs of anxiety, fearfulness, reactivity, or aggression. The dogs, both neutered females, mixed-breed dogs, were 2 and 8 years old (mean

5.0 ± 3.0 SE) and weighed between 3 and 18 kg (mean 10.5 ± 7.5 SE) at the time of the sampling period. Dogs were recruited from the local nonprofit organization "Effetto Palla ONLUS", with the aim of enhancing their socialization and adoption rates [16]. Dogs were subjected to regular health screening and behavioral monitoring by a veterinarian with expertise in animal behavior and welfare. In order to be eligible for participation in the reading programme, the dogs were required to be in perfect clinical health (i.e., free from pain, external and internal parasites, and immunized). These dogs' characteristics, behavior, and welfare during the reading sessions have been described also in more detailed in [17]. Child–animal interaction was limited to verbal contact: no child-initiated contacts with dogs were allowed. Children could only talk to the dog, and they did so by praising her or asking whether she enjoyed the story or was getting bored.


**Table 1.** Age, gender, and diagnosis of Children with ASD involved in the project.

M = Male; F = Female; ASD = Autistic spectrum disorder. <sup>1</sup> According to [12].

#### *2.3. Setting Room*

The two dogs were handled by a female veterinarian expert in animal welfare and behavior, who was familiar with them and was always present during the sessions to guarantee their well-being.

Sessions were performed in a 6 × 5 m carpeted room at the facility, where children were also involved in other activities, in the presence of a psychologist. In more detail, at reading sessions, one visiting dog, one dog handler/veterinarian, one psychologist, and one experimenter were always present. The room temperature ranged between 20 ◦C and 24 ◦C. Two 30 cm high benches were placed to separate the room into two identical spaces, one for the dog and one for the children [10].

#### *2.4. Test*

At baseline (T0) and at the end of the 10 sessions programme (T1), the psychologist administered validated reading and cognitive tests to all the children (Table 2). Pre- and posttest in both groups followed the same order. Reading tests such as Cornoldi reading test (MT2) [18], test of reading comprehension (TORC) [19], metaphonological competence (MCF) [20] and cognitive tests, Wechsler intelligence scale for children (Wisc IV) [21], and Vineland [22].

Session attendance was recorded in both groups. A short self-report questionnaire was prepared by the psychologist by reviewing similar literature [23] in order to collect parents perceptions after the 10 reading sessions. The questionnaire, presented in Table 3, was composed of seven yes/no closed questions. The questions focused on the perception of the parents about: (1) reading motivation, (2) motivation to follow the programme, (3) social skills, and (4) attention towards dogs.

#### *2.5. Statistical Analysis*

Data was analyzed with SPSS, version 25.0 (SPSS Inc, Chicago, IL, USA) through nonparametric statistics as they did not follow a normal distribution (Shapiro–Wilk normality tests, all *p* > 0.05). The Mann–Whitney U test was used to compare differences between the two groups, while the Wilcoxon signed-rank test was used for paired data. Due to the multiple comparisons, Benjamini–Hochberg multiple testing correction [24] was applied. Fisher's exact test was used to investigate associations between the presence of the dog and both children's attendance and parents' answers in the questionnaire. Values of *p* < 0.05 were considered statistically significant.


**Table 2.** Different tests administered to children by psychologist at T0 and T1.

**Table 3.** Parent-completed questionnaire.


EG = Experimental group; CG = Control group.

#### **3. Results**

#### *3.1. Session Attendance*

EG children achieved 100% attendance in each reading session, which was statistically higher than the 75% of CG children (range 25–100%, U = 11.0, z = −3.468, *p* = 0.002, Figure 1). In particular, in CG children, attendance was significantly different on day 9 (Fisher's exact test *p* = 0.04) and day 10 (Fisher's exact test *p* = 0.05) compared with the other days.

**Figure 1.** Children's attendance. EG and CG children. EG = Experimental group; CG = Control Group; *\* p* < 0.05.

#### *3.2. Reading Tests*

We explored the two domains of MT2, namely speed (S) and accuracy (A), the reading comprehension (RC) for the TOR test, and the five domains [18] for the MCF test: Recognition (RE) Fluidity (F), Phonemic (FO), Segmentation (SG), Letter deletion (LD). On all reading tests, no significant differences were found between T0 and T1 within each group (Mann–Whitney U test, *p* > 0.05) or even between the EG and CG groups at each time point (Wilcoxon signed-rank test, *p* > 0.05) (Tables 4 and 5).


**Table 4.** Statistical results of the MT2 and TOR tests at baseline and at the end of the reading sessions (*p* > 0.05). Mean ± Standard Deviation is reported.

MT2-S = Cornoldi reading test speed; MT2-A = Cornoldi reading test accuracy; TOR-RC = TOR test reading comprehension.

**Table 5.** Statistical results of the CMF tests at baseline and at the end of the reading sessions (*p* > 0.05). Mean ± Standard Deviation is reported.


RE = Recognition; F = Fluidity; FO = Phonemic; SG = Segmentation; LD = Letter deletion.

#### *3.3. Cognitive Test: WISC IV Test and Vineland Tests*

The five domains of the WISC IV test have been explored: Intelligence Quotient (IQ), Fluid Reasoning Index (RF), Processing Speed Index (PS), Verbal Comprehension Index (VC), Working Memory Index (WM). Vineland test's domains have been analyzed: Compressive Results (CO), Communication (CM), Daily Living Skills (DLS), Socialization (S), and Motor Skills (MS).

On all the cognitive tests, no significant differences were found between T0 and T1 within each group (Mann–Whitney U test, *p* > 0.05), as well as between the EG and CG groups at each time point (Wilcoxon signed-rank test, *p* > 0.05) (Tables 6 and 7).

**Table 6.** Statistical results of the WISC tests at baseline and the end of the reading sessions (*p* > 0.05). Mean ± Standard Deviation is reported.


IQ = Intelligence Quotient; RF = Fluid Reasoning Index; PS = Processing Speed Index; VC = Verbal Comprehension Index; WM = Working Memory Index.

**Table 7.** Statistical results of the Vineland tests at baseline and the end of the reading sessions (*p* > 0.05). Mean ± Standard Deviation is reported.


CO = Compressive Results; CM = Communication; DLS = Daily Living Skills; S = Socialization; MS = Motor Skills.

#### *3.4. Parents' Questionnaire*

Questions and answers reported by the parents of the EG and CG children are shown in Table 3.

#### **4. Discussion**

In the present study, we evaluated a programme that aims to understand the impact of 10 weekly reading sessions with dogs on children with ASD to read. Attendance and parents' perceptions were evaluated. Validated reading and social tests were employed prior to the beginning and after the end of the programme in order to offer an evidence-based evaluation approach. To our knowledge, this is the first time that tests measuring reading and social skills have been applied to assess the effectiveness of a reading-to-dog programme in children with ASD. Being willing to interact with dogs has been considered as inclusion criteria: this makes it difficult to generalize the results to all children with ASD, although it might be applicable to other children with ASD who happen to like dogs (or at least not dislike them). However, this study wants to work as a pilot in the reading-to-dog programme field.

Motivation has been defined as "a psychological process in which personality traits (e.g., motives, reasons, skills, interests, expectations, and future perspectives) interact with perceived environmental characteristics" [25]. Thus, student motivation can be affected by changes in their learning environment. In our study, the reading-to-dog programme significantly increased the propensity of children to read at home and look autonomously for a book, as showed by EG parents scoring higher on the related questions of the survey compared with CG parents immediately after the end of the programme. This is in line with what is reported in reading studies [1], in which motivation is often discussed in terms of intrinsic motivation (motivated from internal factors; e.g., curiosity to read, enjoyment of the experience) and extrinsic (motivated by external factors; e.g., to get a good grade). Children in the EG group were also perceived by their parents as having a significantly higher motivation to follow reading sessions. EG children actually attended the sessions significantly more frequently (100% attendance) than those in the control group (25% to 100% attendance). According to Newman-Ford et al. [26] attendance is a measure of a student's motivation for learning, which is considered a galvanizing

energy in the learning process. From this perspective, it is not surprising that EG children were significantly more motivated to do homework at home than controls, as reported by their parents. The dog might have acted as motivator for children to attend, which might be due to a dog's recognized ability to be an active, nonjudgmental listener [1]. As reported in [27], "The dogs 'listened' while the students were reading at their own pace. The dogs did not laugh, judge or criticize them, and therefore they were not embarrassed by their own mistakes". Moreover, in [28]), children with autism interacted most frequently and for the longest periods with a real dog in comparison with objects or a person. The presence of the dog assumed an important role during the session. The authors concluded that students reading in the presence of a dog were more likely to participate in reading-to-dog sessions, because that was an environment in which they could build their self-confidence [27]. Also. children with pervasive developmental disorders (including autism) were more playful in interaction with a live dog compared with toys, and also more aware of their social environment in the presence of the dog [29]. However, in the questionnaire, when parents were asked to answer to a specific question about "attention to dogs", no differences between EG and CG were reported. We wanted to ask this question in order to understand if the children with ASD were more aware of the social environment, as reported in literature [29].

In our study, children's engagement in social interactions with peers was not increased at the end of the programme in both groups, according to parents' perceptions. Similarly, Socialization Area results obtained on Vineland tests (for example, the score related to Plays with peer/s for 5 min under supervision, Plays with peer/s for 20 min under supervision, Asks others to play or spend time together) showed no improvements in social skills of children from both groups when the programme was over. This is in contrast with what was reported in the study by Grigore et al., [14], in which the author found improved social interactions in three preschool autistic children following a combined social story method and canine-assisted intervention. As far as we know, there are no other published researches conducted with children with ASD reporting results based on engagement in social interactions with peers. Paul and Serpell [30] found that normal families who obtained a dog, 1 month later engaged in more leisure activities together and their children were more often visited by friends. In a classroom of first-graders, the presence of a dog led to a better social integration among students, as documented via indirect psychometric indicators [31] as well as via direct behavior observation [32].

The possible role of the Oxytocin (OT) in these child–dog interactions during reading-to-dog sessions needs to be underlined too. Nagasawa et al. [33] assessed the effect of 30 min of interaction between dogs and their owners, particularly the duration of friendly gazes from the dogs to the owners. In a control condition lasting for 30 min, owners were instructed not to look at their dogs directly. In the normal interaction condition, longer gaze was linked to higher OT levels in the owner, while this was not the case in the control condition without eye contact. The interaction, even without direct contact, is related to OT increases that are strictly related to social interaction (see [34] for a detailed review). The release of OT via contact with animals may contribute to explain many of the effects of dog–human interactions.

As for both groups, we found no significant gains in children's reading test (MT2, TOR, MCF) scores after taking part in our reading-to-dog programme. In contrast, Konarski et al. [11] reported improved Dynamic Indicators of Early Literacy Skills (DIBELS) and Elementary Reading Attitudes Scale (ERAS) in a child with autism spectrum disorder after completing a reading-to-dog intervention. However, this was a case study, which did not use any control measures or include a case series, and therefore it does not allow to conclude that any change observed is due to the intervention being studied rather than to other factors. Several other authors described positive effects of reading-to-dog programmes in children (see [1] for review). For example, Fisher et al. [35] applied the Neale Analysis of Reading Ability [36] to test reading abilities in one child, before and after participating in a BaRK programme. BaRK is a free programme that involves reluctant readers in the middle-upper primary school classes. In this programme, a child was involved in eight weekly reading sessions with a dog. The results indicated a dramatic improvement between pretest and post-test scores for both

reading accuracy and comprehension, with greater gains being made in comprehension skills. In [37], 26 children had higher scores after reading to a dog on the Gray Oral Reading Test (GORT-4), in which the child has to read aloud narrative passages (of medium length) and, for each passage, answer to multiple-choice comprehension questions read by the examiner. These results were supported by those collected by The Intermountain Therapy Animal [38] that indicated students' reading skills improved by two to four grade levels during a reading programme. However, again, failure to use appropriate controls makes it impossible to draw meaningful conclusions from these studies. Booten [39] and Petersen [40] included a control group in their investigation, and they did not report any differences between children who read to a dog and those who read without a dog. Conversely, Treat et al. [41] found improved reading fluency, accuracy, and comprehension after reading to a dog, while in the study by LeRoux et al. [27] children in the dog group scored higher on the Neale reading comprehension test compared with the control groups, and Kirnan [42] found an improvement in reading skills based on teachers perception. It should be noted that all these studies involved typical children, making it difficult to compare results with ours. In fact, a meta-analysis by Fuchs [43] revealed that the reading achievement of students with a learning disability is significantly different from that of typical students, even if low-achieving students are considered: students with learning disabilities have more severe reading problems than others [43]. Overall, children with ASD can be characterized by a triad of persistent impairments with core deficits in social interaction, language, and communication, as well as restrictive, repetitive thoughts, routines, and behavior patterns: ASD and learning disability are then co-associated. ASD is more likely to be present in individuals with a learning disability, impacting on all aspects of learning, especially among more severely affected individuals [44]. In our study, children with ASD had to follow important but easy rules related to the setting. The reason for this limitation is dual. As for dogs, this allowed activities to be predictable and controllable [17,45]. For children, it was a way to receive a simple but useful rule.

There are some limitations to our study, so the findings should be interpreted with caution. First, the programme involved a small sample size and did not control for the confounding effect of variables, including parenting styles but also comorbid outcomes such as anxiety, which makes it difficult to generalise to a wide population. Second, although parent-completed questionnaires are considered as accurate as developmental screening instruments (see [46] for example), parents were required to interpret their children's motivation and attitudes, inevitably resulting in a degree of subjectivity. In addition, it is possible that the parents' answers were influenced by perceptions of which answers would be deemed acceptable, even if the questionnaire was anonymous. Third, we implemented a short-term intervention, and future studies should examine interventions over a longer time (e.g., the entire school year), possibly analyzing academic performances. However, a standard programme for Children with ASD has not been developed and validated yet [42].

#### **5. Conclusions**

In conclusion, reading to a dog has the potential to bring significant improvements to typical children's social and reading abilities [1]. The results of the present pilot study suggest that such a programme can have specific effects on session attendance and literacy motivation at home in children with ASD, as perceived by their parents. Previous research demonstrated that increased engagement in reading is linked to improved academic performance [47,48]. Thus, the attendance at (and engagement in) reading sessions, enriched by the presence of a dog should be further examined, together with the critical aspects of literacy, including testing accuracy, fluency, and comprehension.

The spatial setting used in this pilot can be applied in following studies in order to create a perfect welfare area for dogs and to take the chance to teach a rule to the children with ASD. In order to evaluate the success of a reading-to-a-dog programme, not only validated tests but also percentage of attendance and parents' perceptions should be taken into account. The next step should include large-scale, randomized control trials with longitudinal examinations of effects, to provide more tangible and reliable findings not only for children with ASD but also for dogs.

A recent review [1] reported positive results based on implementation of a reading-to-dog programme. Unfortunately, these are mostly based on ad-hoc reports, without undergoing a peer-review process [42]. These studies did not randomly allocate children to intervention or control groups, and only small groups (or case study) were investigated. Although extensive generalization should be avoided, the results of our study provide some tentative support for the effectiveness of a reading-to-dog programme based on the use of objective assessments. Specific tests did not confirm any effect on children's social and literacy skills due to the presence of a dog. More research is therefore needed to understand the impact of this type of intervention, considering potential confounding variables, including individual factors or a different number of sessions.

**Author Contributions:** The idea for the paper was conceived by S.U. and F.P. The experimental protocol was designed by S.U. and F.P. and carried out by S.U. and S.F. The data was statistically analyzed by S.U., and discussed by all authors (S.U., M.A., L.P., S.F., F.P.). The paper was written by S.U. and F.P., and revised by all authors (S.U., M.A., L.P., S.F., F.P.).

**Funding:** This project was supported by the Nestlé Purina Sponsorship for Human Animal Bond Studies and by LINEA2FPIRR\_2017 University of Milan (19703\_PSR2015-1716MDI\_G\_M).

**Acknowledgments:** We are grateful to all the children and family who participated in the study. Thanks to Effetto Palla Onlus for providing dogs for the study and Giulia Schirosi for the help in video analysis.

**Conflicts of Interest:** The authors declare no conflict of interest. The funding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.

**Ethics Statement:** The study was performed according to the Declaration of Helsinki. The ethics committee of the CTR Onlus approved the study, protocol number CR3/403-18, 2018. No invasive intervention or drug experimentation on the dogs was performed; therefore, the application of Legislative Decree No 26/2014 and Directive2010/63/EU for the protection of animals used in scientific and experimental studies were not required. The Effetto Palla Onlus provided consent for use of the dogs in the study.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Thyroid and Lipid Status in Guide Dogs During Training: E**ff**ects of Dietary Protein and Fat Content**

#### **Biagina Chiofalo \*, Esterina Fazio, Salvatore Cucinotta and Cristina Cravana**

Department of Veterinary Sciences, University of Messina, Polo Universitario Annunziata, 98168 Messina, Italy **\*** Correspondence: biagina.chiofalo@unime.it; Tel.: +39-0906766833

Received: 18 July 2019; Accepted: 19 August 2019; Published: 23 August 2019

**Simple Summary:** Nutrition is one of the main causes of thyroid response and energetic metabolism. Presently, there is a lack of information on the physiological effect of moderate activities in canines, particularly in guide dogs. Aim was to compare the effect of diet on thyroid and lipid status in guide dogs, during a 12-weeks training, fed two commercial diets, one, HPF, characterized by low-carbohydrate/high-protein/high-fat (29:39:19% as fed) and the other, LPF, characterized by high-carbohydrate/low-protein/low-fat (50:24:12% as fed) content. Our hypothesis was that the intake of a diet rich in fat and protein would have given a better response than the carbohydrate-rich diet for thyroid and lipid homeostasis to cope with the increased energy demands of dogs subjected to the training period. Results evidenced that the consumption of a diet rich in fat and protein appears the nutritional plan most suitable to support moderate exercise for guide dogs during the training work; this diet supports maintenance of body weight, Body Condition Score (BCS), and serum baseline thyroid and lipid profiles, offering potential improvements in dogs' performances. However, the chronic ingestion of diets that are extreme in their composition of either fat or carbohydrate should be always approached with caution.

**Abstract:** Nutrition plays a leading role that most influences thyroid response and energetic metabolism. Aim was to compare the effect of diet on thyroid and lipid status in guide dogs during a 12-weeks training period. Eight Labrador Retrievers were divided into two groups homogeneous for sex, age, body weight, and Body Condition Score (BCS) and fed two commercial diets one, HPF, characterized by low-carbohydrate/high-protein/high-fat (29%:39%:19% as-fed) and the other, LPF, by high-carbohydrate/low-protein/low-fat (50%:24%:12% as-fed) content. The serum thriiodothyronine (T3), thyroxine (T4), cholesterol (CHOL), triglycerides (TAGs) and non-esterified fatty acids (NEFA) were determined at Day 0, 28, 56, and 84, before the daily training. Statistical model included the effects of Diet (HPF vs. LPF) and Time (Day 0 to Day 84), and their interaction. In the HPF group, Diet significantly (*p* < 0.01) increased T4, CHOL, and TAGs and decreased NEFA. In both groups, Time significantly (*p* < 0.05) increased T4 and TAGs, CHOL at Day 28, and NEFA at Day 56. The interaction did not influence serum hormones and lipid pattern. The adjustments in thyroid and lipid responses to moderate exercise in HPF group were driven mainly by the nutrient composition of the diet in relation to the involvement of metabolic homeostasis.

**Keywords:** guide dog; dietary protein/fat ratio; training; thyroid hormones; lipid panel

#### **1. Introduction**

Thyroid hormones are known to play a pivotal role in growth regulation, cognitive issues, cellular function and metabolic implication [1,2]. It was shown that the hypothalamic-pituitary-thyroid (HPT) axis activity decreases in response to food restriction, which is frequently interpreted to be an energy-saving mechanism [3]. It has also been speculated that energy signaling, like obesity and energy restriction, alters thyroid homeostasis in dogs, with significant higher T3 and T4 concentrations in obese dogs than lean dogs [4]. Moreover, the dietary protein quality and quantity content could change HPT axis activity; in particular, the long-term low-protein diet affects the thyroid axis activity, with the effect similar to that caused by starvation [5]. Some studies have shown a correlation between T3 concentrations and resting metabolic requirements [6,7]; nevertheless, no correlation between body weight and serum T3 concentrations was observed [4]. Moreover, serum concentrations of adiponectin were significantly and negatively associated with T4 concentrations, and positively correlated with cholesterol [8]. It is interesting to note that there is also an evidence that thyroid hormones and lipoprotein alterations may have a role in susceptibility of dogs to infectious diseases [9]. Thus, thyroid hormones in dogs may be involved in the regulation of fatty acid delta-6-desaturase activity [10]. The background of the thyroid signal transduction [11] and lipid metabolism in dogs exhibits some unique characteristics compared to other species, and hyperlipidemia is common in dogs [12].

The primary plasma changes, that are needed to support long-lasting activity, are related to lipid metabolism and, for this reason, the lipid pattern is frequently assayed during clinical evaluation [13]. Fatty acids are an important source of energy for skeletal contraction, particularly during exercise of mild-moderate intensity, prolonged duration and in the fasting state [14]. Plasma free fatty acids (FFAs) transported from remote adipose tissue stores and triglycerides (TAGs) contained within skeletal muscle fibers are the major sources of these fatty acids. The relative contribution of each source is dependent on the mode, intensity and duration of exercise and on training status.

Cholesterol arrives in the small intestine from both the diet and bile. The liver –not the diet– is therefore the primary source of cholesterol available for intestinal absorption, a point that is often underappreciated [15].

The relative contribution of fat and carbohydrate changes according to intensity and duration of exercise, the physical training state and the environmental conditions in which the animal is working [16]. With light prolonged exercise, there is a progressively greater use of fat until it can contribute up to 80% of the caloric expenditure. Consumption of a diet rich in fat and protein produces a shift toward a greater use of fat, with a concomitant reduction of both the intensity and duration of effort that can be sustained. Conversely, ingestion of a carbohydrate-rich diet increases the percentage of carbohydrate used and increases endurance [17].

To date, canine athlete physiology studies have primarily focused on endurance sled dog racing and high intensity short duration Greyhound racing. There is a lack of information on the physiological and biochemical changes of low intensity endurance activities in canines, particularly in the guide dog [18]; this information is important in the determination of fitness level, detection of exercise-induced injury and improvement of success, with the development of more specific training programs [19].

Our hypothesis was that the intake of a diet rich in fat and protein would have given a better response than the carbohydrate-rich diet for thyroid and lipid homeostasis to cope with the increased energy demands of dogs subjected to the training protocol.

The aim of this study was to compare the dietary effect of two different concentrations of protein/fat/carbohydrates ratio on total iodothyronines and lipid panel in guide dogs for the blinds (GDB) during a 12-weeks training period.

#### **2. Materials and Methods**

Operative procedures and animal care were carried out in compliance with guidelines of Good Clinical Practices [20] and European regulation [21]. On the basis of the Italian regulation on animal experimentation and ethics [22], the research received the institutional approval by the Ethical Animal Care and Use Committee of the Department of Veterinary Science, of the University of Messina on 19 October 2016, Codex 006/2016 bis.

#### *2.1. Animals and Diets*

The study was carried out on eight neutered adult Labrador Retriever dogs, clinically health, housed at the Regional Centre Helen Keller (www.centrohelenkeller.it)of the Italian Blind and Guide Dog School Union, in Messina (Italy), during the training work to guide service for the blind. The Centre is unique in Italy and it is a member of the International Guide Dog Federation (IGDF), and as such accredited to the highest international standards.

The trial was carried out on dogs without a history of diabetes or hypo- or hyperthyroidism. Dogs admitted to the study were divided into two groups homogeneous for sex (2 male, 2 females), age (HPF: 18.3 months; LPF: 17.5 months), initial BW-body weight (HPF: 26.9 kg; LPF: 26.5 kg) and Body Condition Score-BCS (HPF: 4.33; LPF: 4.5, score 1–9). The first group, called HPF group, received a "performance" diet, characterized by low-carbohydrate/high-protein and fat diet (29%:39%:19% as-fed), whereas, the second group, called LPF group, received a "normal maintenance" diet characterized by high-carbohydrate/ low-protein and fat diet (50%:24%:12% as-fed).

Dogs were individually housed in pens of six square meters, adjacent to a large outdoor space where they could access during rest, and food was administered two times a day, in an individual bowl.

The trial was preceded by 7 days of adaptation period to the experimental diets; anonymous was a normal maintenance diet usually used in the Centre along training work. The adaptation diet was constituted by the mixture of the feeds; the HPF group received a mixture of anonymous with "performance diet" and the LPF group a mixture of anonymous and "normal maintenance diet"; during the 7 days, the anonymous was progressively replaced by the experimental diets. The quantity of administered diet was the same previously adopted by the breeder.

During the trial, the Company sent three lots of feed. Each lot was separately sampled and analyzed, as described by Chiofalo et al. [23].

Both the experimental diets of the Farmina Pet Foods line contained lamb meal as main protein source and, from a qualitative point of view, the same ingredients, analytical compounds, nutritional additives and antioxidants (tocopherol-rich extracts of natural origin). The information on the chemical composition of "performance" and "normal maintenance" diets is reported in the Table 1.

The amount of feed daily administered to each dog was calculated on the ratio between the calculated metabolizable energy requirements, as proposed by Hand et al. [24], for dogs that perform work, characterized by a moderate duration and frequency and the caloric density of metabolizable energy (ME) reported in the label [25], of each diet (HPF and LPF).


**Table 1.** Chemical composition and metabolizable energy of the diets 1.

CP = Crude Protein; OM = Organic Matter; TDF = Total Dietary Fibre; ME = Metabolizable Energy. ND = Not Determined; <sup>1</sup> Values are means ± standard deviation; <sup>2</sup> Anonymous was the normal maintenance diet usually used in the Centre along training work; HPF was the "performance" diet with low-carbohydrate/high-protein/high-fat diet and LPF was the "normal maintenance" diet with high-carbohydrate/low-protein/low-fat diet administered during the trial; <sup>3</sup> Values reported on the label; <sup>4</sup> Values determined analytically.

#### *2.2. Conditioning Protocol*

All dogs were conditioned for training program activities one month prior to the dietary study starts. All dogs were between 1 and 2 years of age and had been conditioned for the training program to be guide dogs in the area described below (see Section 2.3) and they fed a typical maintenance ration (see Section 2.1). Conditioning and training protocols remained the same for all dogs during the dietary trials each of which lasted 12-weeks; thus, each dog served as its own control during the dietary trial. Moreover, the dogs were accustomed to the blood collection since, before the beginning of the trial, hematological and biochemical analyses were carried out on each subject to evaluate their health status. Moreover, before the beginning of the trial, in order to assess the clinical status, all dogs were also submitted to a physical examination [26].

#### *2.3. Training Program*

The method of participant recruitment is described by Lloyd et al. [27].

The training consisted of a various phases program in which the dog gradually learned more guide work. This included leading a person in a straight line, stopping at any change in ground elevation as well as overhead obstacles and obstacle avoidance. Feed rewards were used in the guide dogs for the blind training program as a powerful motivation and reinforcement tool for learning and maintaining desired behavior. During each training session (at Day 0, day 28; Day 56 and Day 84), dogs were introduced to specific guide-work behaviors:


The guide dogs for the blind were trained 3 times a week. Each training session lasted approximately 60 min [23].

#### *2.4. Physical Examination*

To evaluate the performance of the studied dogs, from the Day 0 (start of the administration of the new food) to the Day 84, all dogs weekly underwent to physical examination [26] including: level of consciousness; posture and gait; hydration status; rectal temperature (◦C); pulse rate; respiratory rate and breath character; perfusion indicators.

At the same time, on each animal, the BW and BCS were evaluated.

The determination of BW was measured on fasted animals, in the morning at 8:00 am, by using a digital scale.

BCS was evaluated by assigning a rating scale that ranged from 1 (too thin) to 9 (too heavy) using the table proposed by Nestle Purina [28–30].

#### *2.5. Measurements of Hormonal and Lipid Patterns*

In order to evaluate hormonal and lipid patterns in the fasting dogs, blood samples were monthly withdrawn, at Day 0, Day 28, Day 56, and Day 84 before the exercise (8:00 am).

Before the trial, the dogs were accustomed for the blood collection procedure (see paragraph *Conditioning protocol*). All samples were collected by the same operator into evacuated tubes (Venoject, Terumo®, Shibuya, Tokyo, Japan) and were immediately refrigerated at 4 ◦C after collection; the

samples were subsequently (within 1 h) centrifuged for 15 min at 1500× *g* and collected and stored at −20 ◦C until their analyses. Serum total iodothyronine concentrations were analyzed in duplicate using commercial immunoenzymatic assays (RADIM, Rome, Italy). The method is based on a competitive immunoenzymatic assay and the reagents were prepared as described in the manufacturer's protocol. Total iodothyronines (T3 and T4) in the sample competed with T3 and T4 conjugated with horseradish peroxidase (conjugate) for binding to specific antibody sites of anti-T3 and anti-T4 coated on the wells. At the end of the incubation, all unbound material was removed by aspiration and washing. The enzyme activity which was bound to the solid phase would be inversely proportional to the concentration of T3 and T4 in calibrators and samples, and this was evidenced by incubating the wells with a chromogen solution (tetramethylbenzidine) in substrate buffer. Colorimetric readings were taken using a spectrophotometer at 450 nm (Sirio S, Radim/Seac Co., Rome, Florence, Italy). The sensitivities of the assays were as follows: 0 to12.3 nmol/L for T3, and 0 to 512 nmol/L for T4. The lower detection limits for T3 and T4 were 0.15 nmol/L and 12.8 nmol/L, respectively. The intraassay and interassay variance coefficients were 5.5% and 6.1% for T3 and 4.9% and 8.4% for T4, respectively.

Serum was analyzed for triglycerides (TAGs) using the enzymatic colorimetric method (GPO-PAP, glycerol-3-phosphate oxidase-p-aminophenazone) of McGowan et al. [31], for cholesterol (CHOL) using a modified Abell-Kendall/Levey-Brodie method [32] and for not esterified fatty acids (NEFA) by a coupled enzymatic reaction system (ACS-ACOD Method). First, Acyl CoA Synthetase (ACS) catalyzes fatty acid acylation of coenzyme A. Next, the acyl-CoA product is oxidized by Acyl CoA Oxidase (ACOD), producing hydrogen peroxide which reacts with the kit's Colorimetric Probe. The colorimetric reading was taken using a spectrophotometer at 570 nm.

#### *2.6. Statistical Analyses*

To account for the study design, a mixed model analysis of variance [33] with the fixed effects of Time (Day 0, Day 28; Day 56 and Day 84) and Diet (HPF vs. LPF) was applied. The interaction (Diet × Time) was forced into every model. Random effects in the model were individual dog. Residuals were examined for normality; in each case residuals were normally distributed. Least Squares Means (LSM) and standard error of the mean (SEM) were calculated. The comparison between LSM were performed using the Tukey test. Differences were considered significant for *p* < 0.05.

#### **3. Results**

The effect of environmental temperature is unlikely to play a significant role in this population's energy requirement. It is known that the temperatures outside of the thermoneutral zone of 20 to 30 ◦C will increase the energy requirements by 1 to 5 kcal • BW0.75 per ◦C per day when above or below this zone [25]. During the 3 months of the study (1 March to 24 May), the dogs spent their time in thermoneutral zone (23 ± 2 ◦C); considering the kennel's geographic location, it was unlikely that the temperatures at night dropped below the thermoneutral zone.

#### *3.1. Physical Examination and Body Weight*

During the trial, the dogs presented adequate hydration status and rectal temperature within the physiological ranges (38.4 ◦C ± 0.32). The evaluation of pulse at femoral artery showed physiological characteristics about strength and quality, and the pulse rate within reference ranges for dogs (92 bpm ± 14) [26]. The mean of respiratory rate, determined visually or by auscultation as count either inspirations or expirations, was within the physiological ranges (18–29 ± 3) [26]. Mucous membrane color was pink, capillary refill time was less than 2 s.

The results of the present study regarding dog's performance were published by Chiofalo et al. [23]. Briefly, the diet influenced the animal performances (Table 2) in relation to their different protein, fat and carbohydrate contents, showing a significantly higher BW in the HPF group than those of the LPF group, as well as a significantly higher BCS in the HPF group than those of the LPF group.


**Table 2.** Effect of the diets on BW and BCS of the trial 1.

BW <sup>=</sup> Body weight; BCS <sup>=</sup> Body Condition Score; <sup>1</sup> Values are means (LSM) <sup>±</sup> standard error of the mean (SEM); <sup>2</sup> HPF <sup>=</sup> low-carbohydrate/ high-protein and fat diet; LPF <sup>=</sup> high-carbohydrate/low-protein and fat diet; <sup>3</sup> Probability values for the effects of Diet; a, b Within a row, means with different superscript letter were significantly different (*p* < 0.05).

The BW and the BCS of dogs were monitored weekly during the whole time of the 12-week feeding period. As observed by Chiofalo et al. [23], no significant differences of BW in the dogs of the HPF group from Day 0 (26.9 kg) to Day 84 (25.40 kg) were observed, whereas, the BW in the dogs of the LPF group was affected by the time showing a significant (*p* < 0.05) decrease from the beginning to the end of the trial (Day 0: 26.5 kg; Day 84: 23.44 kg). The interaction Diet × Time showed no significant differences (*p* = 0.270). This could be due to the high variability of the BW in each group during the trial. The trend of the BCS showed no significant differences in relation to the Time (*p* = 0.997) and to the interaction Diet × Time (*p* = 0.991) for the whole trial period.

#### *3.2. Hormonal Response*

As regards the trend of iodothyronine concentrations in relation to the diet (Table 3), after the 12-weeks diet intervention, T3 concentration was significantly not influenced whereas, T4 concentration showed significant higher mean level in HPF group than that observed in LPF group.


**Table 3.** Effect of the diet on serum hormonal and lipid panel concentrations for the whole trial period 1.

T3 = Thriiodothyronine; T4 = Thyroxine; CHOL= Cholesterol; TAGs= Tryglicerides; NEFA= Non Esterificated Fatty Acids; <sup>1</sup> Values are means (LSM) ± standard error of the mean (SEM); <sup>2</sup> HPF <sup>=</sup> low-carbohydrate/high-protein and fat diet; LPF = high-carbohydrate/low-protein and fat diet; <sup>3</sup> Probability values for the effects of Diet; a, b Within a row, means with different superscript letter were significantly different (*p* < 0.05).

In relation to the variable Time (Table 4), the T3 response was not significantly influenced during the trial, whereas T4 concentration was significantly influenced, showing significant lower values at the Day 28 than those observed at Day 0, 56, and 84.

The interaction Diet × Time showed no significant differences for T3 as well as for T4 concentrations.


**Table 4.** Profile of serum hormonal and lipid panel concentrations during the trial 1.

T3 = Thriiodothyronine; T4 = Thyroxine; CHOL= Cholesterol; TAGs= Tryglicerides; NEFA= non-esterified fatty acids. <sup>1</sup> Values are means (LSM) ± standard error of the mean (SEM); <sup>2</sup> HPF <sup>=</sup> low-carbohydrate/ high-protein and fat diet; LPF = high-carbohydrate/low-protein and fat diet; <sup>3</sup> Blood sampling at Day 0, Day 28, Day 56, and Day 84, before the exercise; <sup>4</sup> Probability values for the effects of Diet, Time, and Diet × Time; a,b Within row, means with different superscripts letter were significantly different (*p* < 0.05) due to time; x,y Within column, means with different superscript letter were significantly different (*p* < 0.05) due to diet.

#### *3.3. Lipid Pattern*

As regards the trend of lipid pattern in relation to the Diet (Table 3), after the 12-weeks diet intervention, CHOL and TAGs concentrations showed significant higher mean levels, whereas NEFA levels showed significant lower values in HPF group than those observed in LPF group.

CHOL, TAGs, and NEFA levels were significantly influenced by the Time (Table 4), showing in both groups the highest values at the Day 28 for CHOL and at Day 0 and 28 for TAGs and at Day 56 for NEFA.

The interaction Diet × Time showed no significant differences for CHOL, TAGs, as well as for NEFA concentrations.

#### **4. Discussion**

The primary objective of this study was to examine how normal-weight Labrador dogs respond to HPF and LPF diets during a 12-week GDB training programs and to examine the relationships with the potential changes in circulating THs and lipid panel concentrations. It is reasonable to assume that the single components of diet may induce the metabolic changes, according to workload and performance quality. On these bases, the shift of energy metabolism in a catabolic or anabolic direction during training programs and exercise is characterized by a wide range of metabolic hormones changes, such as total THs, according to lipid parameters. Our hypothesis was that the HPF diet would have given a better answer than the LPF diet for thyroid homeostasis to cope with the increased energy demands of dogs subjected to the training protocol.

Guide dogs for the blind have a great social impact because of their invaluable aid in providing independent mobility to people with visual impairment; their service comes at high cost (approximately 25,000 euros) due to the large amount of resources, housing, husbandry and training, required to train such animals [34]. Furthermore, success rates ranged between 50% and 56% for dogs in training [35] contribute to large production costs. Although the most important skills to train in these dogs are obedience, they also must have an appropriate nutritional plan, in order to support physical fitness. Moreover, also the stress derived from the changes of life style (work and kennels condition) may negatively affect food intake and live weight [24], causing metabolic disorders and some significant modifications in laboratory parameters.

For guide dogs, a normal maintenance diet (crude protein = 20–23%; crude fat = 10%–12%) does not meet the requirements during their training work and the use of large amounts of feed is not recommended. The consumption of the "performance" diet, characterized by low-carbohydrate/ high-protein and fat diet (29%:39%:19% as-fed), seems to be more appropriated for light prolonged exercise than the ingestion of a normal maintenance diet rich in carbohydrates [17], limiting the weight loss in the HPF group, as observed for the dogs of the LPF group (−18%). Nevertheless, all the animals during the trial lost weight; this could be due to the training work for the service guide for the blind. Weight loss is normal in guide dogs during the training, according to the exercise and the life in kennel [23]. Moreover, considering then the Labrador retrievers may be genetically predisposed to obesity and consequently to the osteoarticular diseases [36], and considering the important role of GDB, they always maintain moderate body weight during the training program.

Major depots of fat accumulation are present under the skin (as subcutaneous fat) and they can be readily observed and evaluated in dogs by using a BCS scale as indicator of the fat mass. If dietary energy intake is less than energy need, fat mass and BCS decrease. Conversely, if intake exceeds requirements, fat mass and BCS increase. This could explain our observation regarding a better BCS mean value of HPF group that that observed in LPF group. On the whole, all the animals showed a BCS within the ideal range (score 4 and 5). As reported by Hand et al. [24], a BCS of 2.5–3.5 (on a scale 1−5) is normal for more pets and for many canine athletes; the same authors observed that a much leaner body composition is desirable for some canine athletes. Even small excesses of body fat may represent an unnecessary handicap for working dogs.

In the present study, the circulating T3 and T4, CHOL, TAGs, and NEFA concentrations are reported for the first time in clinically healthy Labrador Retriever guide dogs during training. The comparison of hormonal data with published ranges for dogs revealed that T3 and T4 concentrations were in agreement with physiological wide ranges reported in literature [37,38]. The results of the lipid pattern ranged within the reference values [39] in agility dogs undergo during exercise [19] and in dogs during low intensity endurance activity [18].

Korhonen [40] monitored the levels of THs, total lipids and urea of adult farmed raccoon dogs, and compared these parameters with BW and feed consumption during intense, maintenance and restricted fasting feeding. He observed a marked adjustment of thyroid hormones as the result of changes in subcutaneous fat reserves. This could explain the significant differences observed in our trial for T4 and BCS between HPF and LPF groups, according also to Eshratkhah et al. [41] that reported an influence of THs on lipid metabolism, through increasing lipolysis in adipose tissue and stimulating lipogenesis, by increasing the activities of some enzymes. In fact, the suitable function of thyroid gland is essential to metabolic regulations and for maintenance of the energy balance of body [42]. THs appear to contribute in the body energy balance, modulating the basal metabolic rate, primarily through actions in brain, heart kidney, liver, adipose tissue, and skeletal muscle [43]. The significant differences observed for T4 between the HPF and LPF groups confirm that thyroid hormones were influenced by altering feed intake, changes in subcutaneous fat reserves [40], such as by different diets in relation to the quantity and quality of nutrient contents [44].

The consistent tendency to decrease of the T4 concentrations in LPF diet group could be probably due to the continued weight loss showed by all dogs during the trial. This result confirmed previous studies related to a decrease in T3 and T4 concentrations in dogs undergoing a weight loss protocol [4]. It is well established that thyroid hormone status correlates with BW and energy expenditure [41]. This could be an energy-saving mechanism related to a down regulation of HPT axis activity in condition of caloric deprivation [3]. In human and rats, it's been previously reported that during food or caloric restriction, total body energy expenditure can slow down with an adaptive decrease resulting in a fall in circulating THs [6,7]. Iodothyronines modulate the fat metabolism, and alterations in T4 may reflect increased lipolysis to offset reduced feed intake [2]. The significant decrease in T4 and no change in T3, in response to reducing energy intake, were unexpected. However, our data confirmed other studies that have examined the HPT axis response to starvation with a decrease in Thyroid Releasing Hormone, Thyroid Stimulating Hormone-β gene expression and circulating THs in rodents and human [3]; the decreased T3 may be primarily due to diminished thyroidal secretion of T4 or by increased Desiodinase (D3) activity in the liver, kidney, and muscles of starved rats [45].

A particular note should be done on the thyroid response to variable diet. Only at the last of the trial (Day 56 to Day 84), data showed a significant higher basal concentration of T4 in HPF group than the mean value of LPF group. The positive value trend might relate to a reduced catabolism resulting from decrease lipoprotein lipase activity [46], which determined a positive energetic balance of the animals fed the "performance" diet rich in fat and protein.

These results could indicate that adjustments in thyroid function and related consistent increase of circulating T4 concentrations in HPF group, that were driven mainly by the nutrient composition of the diet in relation to the involvement of THs in the synthesis, mobilization, and degradation of lipids [47].

As TAGs are the most important type of fat in the diet and the body's primary for stored energy, during prolonged exercise and when energy intake is insufficient, they are metabolized in FFAs determining an increase of NEFA in the blood which became a primary energy source for long-lasting exercise [48]. Plasma FFA oxidation is directly related to the rate of lipolysis in adipose tissue [14]. Their oxidation can contribute 50 to 60 per cent of the energy expenditure during a bout of low intensity exercise of long duration [49]. This could confirm the better BCS and the lower NEFA levels recorded in the HPF than LPF groups related to the more adequate energy content of the "performances" diet than that of the "normal maintenance" carbohydrate-rich diet. Moreover, the results of this study indicate that the diet induces significant changes in TAGs concentrations. Circulating TAGs could be a potential source of fatty acids for ß-oxidation in working muscle, especially in animals in the fed state [50]. The rise in TAGs concentration after exercise depends on the intensity of exercise and the activity of lipolysis, although FFA concentrations are considered to be a better indicator of lipid metabolism [51]. The "normal maintenance" diet, characterized by high-carbohydrate/low-protein and fat diet (50%:24%:12% as-fed), increased blood TAGs, as effect of lipolysis stimulation inducted from the high request of energy during the metabolic adaptations that occur in skeletal muscle and adipose tissue, and that facilitate a greater delivery and oxidation of fatty acids during exercise. Our results are in accordance with Askew [49] and Kaciuba-U´scilko et al. [52] that observed a markedly enhancement of FFA mobilization modulated by the thyroid hormones, in relation to a decreased feed intake.

Total CHOL concentration is routinely measured during health checks in small animal clinics [13]. The total serum concentration of CHOL has been recognized as a potential biomarker for various processes related to lipoproteins metabolism [15]. Fialkoviˇcová et al. [53] reported that THs have catabolic effects on muscle and adipose tissue and regulate CHOL synthesis and degradation; they are essential for an appropriate degree of metabolic activity, including generation and release of energy. It is possible presume that the higher CHOL levels of the HPF than LPF groups, would be probably correlated to a greater intestinal absorption of medium and long chain fatty acids, which would be esterified in situ and introduced again as lipoproteins and chylomicrons into the blood, testifying to an improvement in the intestinal absorption of the nutrients of HPF group.

Our data are not in accordance with Bruss' observations [54] where serum CHOL level generally varies inversely with thyroid activity. However, there are some contradictory findings regarding the relation between serum THs, CHOL, and TAGs; the concentrations of THs were not correlated with CHOL levels in some other animals [41,55,56]. Although the role of thyroid hormones is well known in many species, there is little evidence describing the relationship between thyroid hormones status and serum profiles of CHOL, TAGs, and NEFA in dogs [57].

This could probably due to the daily rhythmicity of total lipids, total CHOL, phospholipids, and TAGs that occurs in some animals and that vanished when dogs were food-deprived, indicating that these rhythms are driven by the digestive process [58].

Finally, literature data [54,57] report that changes in concentrations of THs in some animal species are due to the effect of temperature and season. Our trial was carried out in spring and the dogs spent their time in thermoneutral zone therefore, we think the environmental temperature could not have influenced the energy requirements, the functional activity of the thyroid gland and the concentration of THs [57].

On the whole, the significant higher values of BW, BCS, TAGs, and CHOL, together with T4 concentrations, and the significant lower NEFA concentrations in the HPF group, testified a better physical fitness of the animals fed the "performance" diet [59].

#### **5. Conclusions**

Clinical biochemistry parameters are of major interest in canine sport medicine to assess health status and fitness level, as well to monitor the mental and physical stress imposed by exercise. Presently, there is a lack of information on the physiological effect of moderate activities in canines, particularly in the guide dog.

Furthermore, for working dogs, the lifetime cost of feed, even if specially formulated, represents a trivial fraction of the monetary investment in training. It is worth noting that guide dogs for the blind are expensive to train, as well as being expensive in personal terms for all concerning if the post-qualification period is unsuccessful; thus, this research is intrinsically leading for the guide dog trade in several ways. The knowledge of metabolic changes is essential in order to design specific and individual training protocols, for an early diagnosis of poor performances, to assess the impact of different feeding or supplementation strategies and to minimize the risk of exercise-linked disease. Results evidenced that the consumption of a diet rich in fat and protein appears the nutritional plan most suitable to support moderate exercise for guide dogs during their training work, supports maintenance of BW, BCS, and serum baseline thyroid and lipid profiles, contrasting the mobilization of subcutaneous fat reserves, and offers potential improvements in challenging work situations. However, the chronic ingestion of diets that are extreme in their composition of either fat or carbohydrate should be always approached with caution.

**Author Contributions:** Conceptualization, B.C.; Data curation, E.F. and C.C.; Formal analysis, S.C. and C.C.; Funding acquisition, B.C.; Methodology, S.C. and C.C.; Project administration, B.C.; Writing—review and editing, B.C. and E.F.

**Funding:** This research was funded by the Farmina Pet Foods/Russo Mangimi S.p.A. (Via Nazionale delle Puglie, 80035 Nola-NA, Italy) grant, 2016.

**Acknowledgments:** Authors want to thank for the continuing support the President and the Guide Dog Trainers of the Regional Centre Helen Keller of the Italian Blind and Guide Dog School Union in Messina.

**Conflicts of Interest:** The authors declare no conflict of interest. The funder had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Local Dot Motion, Not Global Configuration, Determines Dogs' Preference for Point-Light Displays**

#### **Carla J. Eatherington 1, Lieta Marinelli 1,\*, Miina Lõoke 1, Luca Battaglini <sup>2</sup> and Paolo Mongillo <sup>1</sup>**


Received: 26 July 2019; Accepted: 3 September 2019; Published: 6 September 2019

**Simple Summary:** Animal motion is characterised by predictable kinematics according to their body morphology and the laws of gravity. This pattern of movement, called biological motion, is traditionally studied using animated displays created by placing a small number of light dots on the major joints of living beings. Previous studies have shown that several animal species can reliably discriminate dot displays depicting an animal walking, and their performance is impeded when the display is turned upside-down and is variably affected when each dot is displaced to disrupt the global biological arrangement. In this study, we investigated this phenomenon in dogs during the presentation of dot displays depicting humans or dogs walking. Our findings showed that dogs preferred to view the display which depicted an upright dog, regardless of its global arrangement, and had no significant preferences when displays depicting humans were presented. This suggests that dogs' sensitivity to biological motion depends mainly on the presence of dot motion that moves in accordance with gravity. Also, our findings suggest that, despite dogs' extensive exposure to human motion, they are not sensitive to the bipedal motion presented in the human dot displays.

**Abstract:** Visual perception remains an understudied area of dog cognition, particularly the perception of biological motion where the small amount of previous research has created an unclear impression regarding dogs' visual preference towards different types of point-light displays. To date, no thorough investigation has been conducted regarding which aspects of the motion contained in point-light displays attract dogs. To test this, pet dogs (*N* = 48) were presented with pairs of point-light displays with systematic manipulation of motion features (i.e., upright or inverted orientation, coherent or scrambled configuration, human or dog species). Results revealed a significant effect of inversion, with dogs directing significantly longer looking time towards upright than inverted dog point-light displays; no effect was found for scrambling or the scrambling-inversion interaction. No looking time bias was found when dogs were presented with human point-light displays, regardless of their orientation or configuration. The results of the current study imply that dogs' visual preference is driven by the motion of individual dots in accordance with gravity, rather than the point-light display's global arrangement, regardless their long exposure to human motion.

**Keywords:** dog; biological motion; point-light display; visual perception; experience

#### **1. Introduction**

Animal motion is characterised by predictable kinematics according to their body morphology and the laws of gravity. Johansson [1] captured this movement by placing a small number of point-lights on the major joints of a human body and found that when viewed in isolation they still created the impression of a moving person—despite the lack of other visual information. Biological motion perception has been extensively researched in humans, the results of which demonstrate that people are able to extract a wealth of information from point-light displays, including gender [2,3], emotional state [4], familiarity [5,6] and action performed [7,8]. The perception of biological motion is also relevant to non-human animals, although instead of assessing their ability to infer specific information from point-light displays, research has tended to focus on demonstrating the relevance of biological motion cues to the species under investigation via conditioned discrimination (Baboons [9]; Bottlenose dolphins [10]; Cats [11]; Chimpanzees [12]; Pigeons [13,14]; Rats [15]) or spontaneous preference tasks (e.g., Dogs [16,17]; Chicks [18–21]; Medaka fish [22]; Marmosets [23]; Mice [24]).

In order to assess an individual's preference towards biological or non-biological motion, point-light displays are often presented in conjunction with manipulated ones. A common stimulus manipulation is to flip the point-light display along the horizontal axis. Inverting a point-light display preserves the spatial relationship between dots, but alters the movement of individual dots, which no longer conform to the laws of gravity. Studies conducted in human infants [25] and visually naive chicks [20] have revealed that this disruption of local dot motion reduces the attractiveness of point-light displays, so they are viewed for less time. It has also been shown that inverting a point-light display impaired cats' [11], marmosets' [23] and pigeons' [14] ability to discriminate biological motion. This influence is particularly apparent in dots representing wrists or ankles, which are therefore believed to represent crucial cues for detecting biological motion [26].

A different common manipulation is to scramble point-light displays, by moving individual dots to a different starting position. Scrambling a point-light display disrupts the spatial relationship between dots but maintains the trajectory and accordance with gravity of local dot motions. The impact of disrupting the display's global structure whilst preserving the local dots motion is less clear and may be dependent on the species under investigation. For instance, research in human infants [25] and chicks [19] revealed no visual preference towards coherent or scrambled point-light displays. On the other hand, mice [24], female marmosets [23] and female chicks [18] looked significantly less towards scrambled point-light displays than coherent displays. Blake [11] also found that scrambling a point-light display impaired cats' ability to discriminate biological motion. However, Parron and co-authors [9] found a higher rate of discrimination transfer between upright coherent point-light displays to scrambled point-light displays than from upright coherent point-light displays to inverted coherent point-light displays. Also, the finding that the inversion effect can still be detected when displays are scrambled [26] suggests that the global structure of a point-light display may be less important for biological motion perception than the motion of individual dots in accordance with gravity.

To date, limited investigation into biological motion perception in dogs has been conducted. The first to present animated point-light displays to dogs were Kovács and co-authors [16], who investigated the role of oxytocin on dogs' sensitivity to human motion. In their study, pairs of stimuli were presented comprising an upright coherent human and an inverted and scrambled human, with or without a background of random dots. The experiment revealed a significant effect of oxytocin in modulating dogs' looking preference, when the point-light displays were not masked by random dots, implying that reducing dogs' responsivity influenced their preference for different types of point-light displays. However, no direct comparison of dogs' visual preference to either stimulus was presented, thus leaving unanswered the question of whether dogs do show a visual preference bias towards human biological motion.

Ishikawa and co-authors [17] investigated the role of sociability on dogs' preference for viewing conspecific and human point-light displays. Several combinations of stimuli pairs were presented, varying in terms of manipulation (upright or inverted orientation), direction of movement (frontal or lateral) and species (dog or human). Dogs' level of sociability towards humans and dogs was measured via a questionnaire completed by the owner, which allowed researchers to categorise dogs into high or low sociability groups. A complex pattern of results revealed that overall dogs looked significantly more at human upright frontal point-light displays compared to their inverted control. However, they also found that high-sociability dogs preferentially viewed human inverted point-light displays when presented in the lateral orientation compared to its upright counterpart. And finally,

that although low sociability dogs preferentially orientated towards upright dog displays presented laterally compared with its frontally orientated control, high-sociability dogs exhibited the diametrically opposite pattern of results.

In summary, the two previous studies into biological motion perception in dogs [16,17] were not able to clearly answer to what types of point-light displays dogs are preferentially attracted. A possible reason for this is because different types of point-light displays contain one or more different motion features (e.g., upright, inverted, coherent, scrambled). Consequently, the aim of the current experiment was to better understand what features of point-light displays dogs preferentially view, by systematically manipulating physical aspects of point-light displays representing both dogs' and humans' motion.

#### **2. Materials and Methods**

#### *2.1. Subjects*

Forty-eight dog-owner dyads were recruited through the database of volunteers at the Laboratory of Applied Ethology in the University of Padua. Twenty-eight dogs were pure-breeds (4 Australian Shepherds, 4 Border Collies, 3 Cocker Spaniels, 3 German Shepherds, 3 Golden Retrievers, 2Weimaraners, 1 Basenji, 1 Czechoslovakian Wolfdog, 1 Dogue de Bordeaux, 1 English Setter, 1 Greyhound, 1 Staffordshire Bull Terrier, 1 Standard Poodle, 1 Vizsla, 1 Whippet) and 20 were mixed-breed dogs (4 small, ≤35 cm at the withers; 10 medium, >35 and <55 cm; 6 large, ≥55 cm). The sample consisted of 29 females and 19 males (mean age ± SD: 5.3 ± 2.6 years). The criteria for recruitment were that dogs had lived with their current owner for the last six months and that they were in good health condition. The study was conducted in accordance with relevant legislation about research involving animals, and, for the type of procedures involved, no formal ethical approval was required.

#### *2.2. Stimuli*

The stimuli consisted of white point-light displays representing walking humans or dogs on a black background (Figure 1). The point-light displays were created by video recording one male and one female from each species walking with lateral orientation from left-to-right or right-to-left at a constant speed for one complete cycle of their legs. Markers were placed on the following joints: atlas-occipital, shoulder, elbow, wrist, hip, knee, ankle and, for dogs only, the metatarsophalangeal and metacarpophalangeal joints. Videos were recorded at 120 frames per second. The videos were stabilised using Adobe After Effects CC 2017 (Version 14.2.1, Adobe Inc., San Jose, CA, USA), so they looked as if the person/dog was walking on a treadmill. The resultant movie clip was imported into Tracker [27], where the coordinates for each joint marker were recorded frame-by-frame. Using these coordinates, point-light animations were created using the BioMotion Toolbox [28] for Matlab (Mathworks Inc., Natick, MA, USA). The resulting animation was looped to create 15 s presentations with continuous motion. The BioMotion Toolbox was also used to create inverted and scrambled versions of the original point-light displays. In inverted versions the overall point-light displays were flipped upside-down, so that the spatial relationships between individual points were maintained, but the characteristics of their local motion were opposed to that of a biological entity with respect to gravity. In scrambled versions each individual dot composing the point-light display was randomly displaced to a different starting position compared to its original location, thereby disrupting the global coherence of the point-light display, whilst maintaining the characteristics of the local motion with respect to gravity. Inversion and scrambling could be combined to obtain inverted-scrambled point-light displays, so four different types of stimuli were created for both human and dog stimuli: upright coherent (UC), inverted coherent (IC), upright scrambled (US) and inverted scrambled (IS) (see Figure 1). Moreover, two versions of each stimulus were created, one where the animated figure appeared to be facing right and one facing left.

**Figure 1.** Screenshots exemplifying different types of stimuli used in the experiment: (**a**) human upright coherent, (**b**) human inverted coherent, (**c**) human upright scrambled, (**d**) human inverted scrambled, (**e**) dog upright coherent, (**f**) dog inverted coherent, (**g**) dog upright scrambled and (**h**) dog inverted scrambled.

#### *2.3. Experimental Setting*

The experiment was conducted in a quiet, dimly lit (approximately 4 cd/m2) room (4.7 <sup>×</sup> 5.8 m) with a large plastic screen (2.4 × 3.4 m) at one end and a Toshiba TDP T100 projector (Toshiba corporation, Tokyo, Japan) mounted 2.15 m high on the wall opposite. Pairs of stimuli were projected onto the screen simultaneously. Human stimuli were sized approximately 130 × 80 cm and dog stimuli were sized approximately 65 × 100 cm. The distance between the centre of each point-light display was 1.80 m. During testing, dogs faced the screen at a distance of 1.65 m, either standing or sitting in between their owner's legs who was seated on a small stool behind them (Figure 2). Owners were instructed to gently hold the dog in place but look straight ahead so as not to influence the dog's behaviour. Trial presentation was controlled by an experimenter seated at the back of the room, using a MacBook Pro. A Canon XA20 (Canon, Tokyo, Japan) camcorder was mounted at floor level, 10 cm in front of the screen and facing the dog's head, to record dog's eye movements. Finally, two CCTV cameras were mounted on the ceiling, facing down towards the dog to record its head orientation.

**Figure 2.** A video-still of the experimental setting, during a presentation.

#### *2.4. Experimental Design*

For each species, all possible combinations of manipulations were paired, therefore obtaining six different trial types per species, that displayed two different stimuli simultaneously (see Table 1). The two stimuli within each trial type were facing the same direction. To contain experimental subject habituation, each dog was only presented with three trial types per species (three dog point-light displays and three human point-light displays), totalling six trials per dog; in addition, the direction of movement of the figures was randomised and counterbalanced among the six trials. Overall, each trial type, for both dog and human point-light displays, was seen by 24 subjects. The presentation order of the six trials was pseudo-randomised, to ensure that each trial type was presented equally as often first, second, third, fourth, fifth or sixth, and that human and dog trials were presented in alternate fashion. Also, the side of presentation of the two stimuli, and the direction of movement of the figures, were counterbalanced across the dogs' sample.

**Table 1.** Combinations of stimuli presented in the six different trial types. Trials featuring these stimuli were created for both dog and human point-light displays. UC = upright coherent, IC = inverted coherent, US = upright scrambled, IS = inverted coherent.


#### *2.5. Test Procedure*

Dogs were initially given ten minutes to become familiar with environment, including the experimenter. Before each trial, dogs were led into the testing room and positioned in front of the screen with their two left and two right paws either side of a central line marked on the floor. Each trial was started when dog was looking straight forward towards the presentation screen; and if the dog did not express the behaviour spontaneously, then their attention was captured by moving the projected computer mouse. At the start of the trial, the two point-light displays composing the trial-type were projected, and held on for 15 s, after which the stimuli disappeared, and a black screen appeared. Dogs were led out of the testing room at the end of each trial, and after a rest period of five minutes they were reintroduced for the start of a new trial.

#### *2.6. Data Collection and Analysis*

Using Observer XT software (version 12.5, Noldus, Groeningen, The Netherlands) a continuous sampling technique was used to collect data about dogs' visual orientation from the videos recorded during testing. Dogs' visual orientation was coded as "left" if the dog was looking at the point-light display to the dog's left, "right" if they were looking at the point-light display to the dog's right, and "elsewhere" if the dog was looking anywhere else in the room. If at any time it was not possible to tell where a dog was looking by the frontal video, then head orientation (videos from above) could be used but this was rarely needed. Inter-observer reliability was assessed using data collected by a second observer for dogs' visual orientation on 20% of videos and was revealed to be good (Pearson's *r* = 0.85). Data collected about the dogs' orientation were used to compute the total amount of time in which dogs looked at the stimuli as well the total amount of time the dogs looked at either stimulus. For the analysis, only data for dogs who looked at the display for a minimum of 5 s were included.

A generalised estimating equation (GEE) model was used to assess the influence of various physical characteristics of point-light displays on the amount of looking time dogs directed towards motion displays. In building the model, being scrambled (yes/no) and/or inverted (yes/no) were included as fixed factors, as were their first-order interactions. The dog's ID was included as a random factor, to account for the repeated sampling from each dog. Bonferroni-corrected post-hoc comparisons were performed when a significant effect was found for any of the factors included in the model.

As humans and dogs were never presented in the same display, the "species" factor was not included in the GEE model described above, and separate models were run on data collected from trials where dogs and where humans were presented. However, an additional analysis was performed on the total amount of attention paid by dogs to either stimulus to determine whether the presentation of dogs or humans had an overall effect in attracting dogs' attention. To this aim, a GEE model was run on total attention as dependent variable, the species displayed as a two-level factor, including the dog's ID as random factor.

All statistical analyses were conducted using SPSS (version 24, IBM, Armonk, NY, USA), with statistical significance level set at 0.05.

#### **3. Results**

An average of 15 dogs in each trial across all trial types (min: 10; max: 19) looked at the stimuli for more than 5.0 s. These dogs directed a minimum of 5.0 s towards both point-light displays, and a maximum of 15.0 s, with a mean ± SD of 9.6 ± 2.9 s, with no significant difference between displays showing dogs or humans (Wald χ<sup>2</sup> = 0.277, *p* = 0.599). Of this looking time, dogs directed a minimum of 0.0 s towards each stimulus, and a maximum of 15.0 s (mean ± SD: 4.8 ± 4.2 s).

#### *E*ff*ect of Stimulus Properties on Looking Time*

Results of the GEE indicating the effect of factors influencing dogs' looking time towards human point-light displays are summarised in Table 2. With regards to human trials, there was no effect of inversion, scrambling or an interaction between the two at either time point. Results of the GEE indicating the effect of factors influencing dogs' looking time towards dog point-light displays are summarised in Table 3. No effect was found on dogs' looking time for scrambling and the interaction between scrambling and inversion at either time point. Conversely, a significant effect of inversion was found, with dogs preferentially looking at upright dog point-light displays (mean ± SE: 5.5 ± 0.4 s; 95% CI: (4.7, 6.2)) compared to inverted dog point-light displays (4.2 ± 0.3 s; 95% CI: (3.5, 4.8); mean difference ± SE: 1.2 ± 0.6 s; 95% CI: (0.1, 2.4)). Mean ± SD looking time towards dog and human point-light displays with different manipulations are presented in Figure 3.

**Table 2.** Results of the Generalized Estimation Equation model on looking time to each stimulus during human trials. df = degrees of freedom.


**Table 3.** Results of the Generalized Estimation Equation model on looking time to each stimulus during dog trials. df = degrees of freedom.


**Figure 3.** Mean ± SD amount of attention paid by dogs to upright-coherent (UC), upright-scrambled (US), inverted-coherent (IC) and inverted-scrambled (IS) light-point figures representing a walking dog or a walking human.

#### **4. Discussion**

The current study investigated which features of biological motion the dogs' directed more looking time towards, by presenting them with pairs of point-light displays of walking dogs or humans that contained aspects methodically manipulated (coherent/scrambled, upright/inverted). The results revealed that dogs directed significantly longer viewing times towards upright dog point-light displays, regardless of their global configuration. No bias in visual preference was observed when dogs were presented with any of the human point-light displays.

The finding that dogs significantly biased their looking time towards dog upright point-light displays corroborates with previous research which found that inverting point-light displays reduced the amount of looking time they attracted (e.g., chicks [20]; marmosets [23]) and impaired visual task performance (cats [11]; pigeons [14]). Traditionally, inversion was believed to impact an individual's ability to process stimuli holistically, a renowned effect observed in face processing [29,30]. However, in the case of biological motion stimuli, the effect of inversion is still present, even when viewing scrambled point-light displays [26], when holistic perception would not be possible. More relevant to the perception of biological motion is that inversion alters the kinematic properties of the moving dots, which no longer move in accordance with the laws of gravity. Thus, a detrimental effect of inversion on viewing times indicates that accordance with gravity is crucial to the detection of biological motion, as observed in chicks and humans [20,31]. This seems also to be the case for dogs in our experiment.

Nevertheless, accordance with gravity as such is not sufficient to explain why our dogs showed a bias in looking time for upright dog point-light displays, but not for upright human ones. A first explanation could be focused on the movement of specific dots. Particularly, human and dog stimuli differed in the amount of limb motion they contained. Not only was this due to the fact that humans are bipeds, but even more feet motion was present in dogs' point-light displays since two joint-dots were marked on every dog limb (ankle and metatarsophalangeal or metacarpophalangeal joints), whilst only one joint was marked on every human leg (ankle). Previous evidence showed that some point-lights provided more movement information than others, implying that there was something "special" about the motion of these dots. For example, Mather and co-authors [32] found that people selected the direction with an accuracy rate of 90% if the shoulder and hip dots, or elbow and knee dots were removed, but performed at near-chance levels if the wrist and ankle dots were removed. It could be argued that the arm movement of the human point-light display contains all the same biological movement that is contained in the legs, rising and falling under the influence of gravity. However, the kinematics of arms and limbs/legs movements are quite different. In accordance with this, Troje and Westhoff [26] also found that local feet motion was crucial for human participants to extract directional information from point-light displays, and based on this they suggested that the local motion contained in animals' feet was used as part of an evolutionary system for detecting animals within their visual environment. This idea is supported by Chang and Troje [33] who claim that it is the vertical acceleration pattern which feet motion contains that is essential to allow the visual system to identity an animal. Also, an electroencephalographic (EEG) study by Wang and co-authors [34] found that humans automatically responded to the characteristics of the local biological motion, but not of the global configuration.

A further aspect that needs to be considered when discussing the lack of effects of scrambling, is ambient luminance. An earlier study showed that humans have difficulty in discriminating scrambled from unscrambled biological motion figures at very low light levels [35]. The authors argued that such conditions only affect the perception of local motion to a limited extent but make it more difficult to assemble local signals into a global percept. Whether this is true, and at what light intensities this occurs, is a matter of debate. For instance, Burton and collaborators [36] report an impairment in the perception of (global) biological motion only in the scotopic range (e.g., when only rod photoceptors are active). Conversely, Billino and collaborators [37] report the greatest impairment in the mesopic range, attributing it to the conflicting, simultaneous activation of the rod and cone systems. Our experiment was conducted at light levels slightly above the threshold between the mesopic and photopic range of humans [38], which would predict no detrimental effect on our dogs' perceptual abilities. However, substantial differences exist between dogs' and humans' vision, including lower acuity at various luminance levels [39], and higher light sensitivity [40], which suggest that thresholds between the photopic, mesopic and scotopic range may also differ between the two species. In this sense, we cannot

exclude that the relatively low ambient luminance contributed to the irrelevance of global configuration in driving dogs' attention to biological motion stimuli.

Another hypothesis that could explain the difference in looking time for upright dog point-light displays, but not for upright human ones might be related to different patterns of neural activity which underlie visual processing of conspecific and heterospecific movements. Previous studies in humans [41,42] and monkeys [43–46] indicate that visual sensitivity to actions depends on the observer's past motor experience with the action being observed. Using functional magnetic resonance imaging, Saygin and co-authors [47] found in humans that brain regions containing mirror-neurons are activated by viewing human point-light displays. Similar findings were also revealed during EEG studies with point-light displays showing different human actions [48] and emotions [49]. In line with these findings, Pinto and Shiffrar [50] found that people demonstrated greater visual sensitivity to coherent human motion than coherent horse motion perception in point-light displays. Thus far, mirror neurons have not been explicitly identified in dogs, but their presence can be assumed on the basis of dogs' performance during experiments investigating emotional contagion and mimicry (e.g., [51]). It is then possible that human motion is not able to attract attention in dogs because it does not activate their mirror neurons since they cannot physically preform the action themselves.

Our results support the hypothesis that the characteristic motion of the limbs of an animal in locomotion are crucial for biological motion detection [26], although detection in dogs may be selectively tuned to other quadrupeds. Whilst the evolutionary usefulness of such capacity is clear, allowing dogs to quickly identify other animals within their environment and act on that information accordingly, it also suggests that neither history of domestication, nor adult pet dogs' extensive exposure to humans, enhanced the salience of human bipedal motion in the same way. However, it remains unclear whether dogs recognised point-light displays as representing other dogs or at least a generic quadrupedal organism, or only perceived the characteristics of the local motion contained within the display. A study by Vallortigara and co-authors [19] suggests that chicks are not able to infer species information from movement, since they spontaneously approached point-light displays of predators as well as conspecifics. Further support for the idea that animals do not recognise the identity of point-light displays was provided by research in human infants [25] and chicks [20] which revealed no looking time bias towards upright coherent or upright scrambled/random point-light displays. The lack of any effect of scrambling in our experiment suggests that dogs may also not be able to recognise dogs in these types of stimuli.

At first glance, our results stand at odds with some previous findings. For instance, Kovacs and co-authors [16] suggested that dogs' preference for unmasked human biological motion was reduced by the administration of oxytocin; however, no main effect for biological motion preference was found in the placebo group dogs which falls in line with what we observed during human point-light trials. Ishikawa and co-authors [17] reported an overall looking time bias towards upright humans compared to inverted point-light displays, which clashes with the lack of inversion effect for humans in our study. However, this finding was observed when the human point-light displays were presented in the frontal orientation which the current experiment did not include. When constrained to dog point-light displays, Ishikawa and co-authors found no significant preferences for lateral upright compared to inverted dog point-light displays. A possible explanation for the lack of any preference found in their work could be the difference in the procedure. Unlike Ishikawa et al.'s experiment where the point-light displays were presented for 5 s, the current study measured dogs' looking time at 15 s time points allowing to capture more sustained looking time biases.

#### **5. Conclusions**

A systematic investigation of dogs' looking time allocation towards different features of point-light displays revealed that dogs did not show any bias towards features of human point-light displays, but preferentially viewed upright biological motion displays originated from walking dogs, compared to inverted ones and regardless of the global configuration. In line with previous research in animals, the finding champions the importance of limb motion in accordance with the laws of gravity for the detection of moving biological entities by dogs.

In spite of some discrepancies between the present findings and those of other studies in dogs, ours and previous research do converge on one remarkable aspect: that the extent of dog's attentional biases towards biological motion displays is rather limited. Considering that visual preference to biological motion seems well conserved across taxa, it seems unlikely that such weak bias merely reflects a scarce sensitivity of the species towards biological motion. Rather, other factors may have contributed to drive dogs' attention towards visual stimuli; for instance, both the novelty of a specific stimulus, as well as the familiarity of the subject with a certain category of stimuli, can interact in determining dogs' preferential looking at certain visual displays [52]. Thus, in ours, as well as in past research, these factors may have been competing to some extent with the attractiveness of figures depicting biological motion.

**Author Contributions:** Conceptualization, L.M. and P.M.; methodology, L.B.; formal analysis, P.M.; investigation, C.J.E. and M.L.; data curation, C.J.E.; writing—original draft preparation, C.J.E.; writing—review and editing, L.M. and P.M.

**Funding:** C.J.E. was supported by a post-doc grant from the University of Padua (Grant Nr. BIRD178748/17), M.L. was supported by a PhD grant from Fondazione Cariparo, and APC were funded by the PhD School in Veterinary Science, University of Padua.

**Acknowledgments:** We are very grateful to Carlo Poltronieri for his technical assistance, to Alessia Cocco, Stephanie Massaglia and Martina Violo for their help running the experiments, and to all the dogs' owners for volunteering their time.

**Conflicts of Interest:** The authors declare that they have no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

## **Examining How Dog 'Acquisition' A**ff**ects Physical Activity and Psychosocial Well-Being: Findings from the BuddyStudy Pilot Trial**

### **Katie Potter 1,\*, Jessica E. Teng 2, Brittany Masteller 3, Caitlin Rajala <sup>1</sup> and Laura B. Balzer <sup>4</sup>**


Received: 7 August 2019; Accepted: 4 September 2019; Published: 7 September 2019

**Simple Summary:** Dog owners are more physically active than non-dog owners, but the direction of the relationship between dog ownership and increased physical activity is unknown. In other words, it is unclear whether acquiring a dog causes a person to become more active, or whether more physically active people choose to acquire dogs. Given that regular physical activity is critical for the prevention and management of numerous chronic diseases, research supporting the hypothesis that dogs make people more active could inform programs and policies that encourage responsible dog ownership. In the BuddyStudy, we used dog fostering to mimic dog acquisition, and examined how taking a dog into one's home affected physical activity and psychosocial well-being. Nearly half of study participants saw large increases in physical activity and nearly three-quarters saw improvements in mood after fostering for six weeks. More than half met someone new in their neighborhood because of their foster dog. Most participants adopted their foster dog after the six-week foster period, and some maintained improvements in physical activity and well-being at 12 weeks. The results of this pilot study are promising and warrant a larger investigation.

**Abstract:** Dog owners are more physically active than non-dog owners, but evidence of a causal relationship between dog acquisition and increased physical activity is lacking. Such evidence could inform programs and policies that encourage responsible dog ownership. Randomized controlled trials are the 'gold standard' for determining causation, but they are prohibited in this area due to ethical concerns. In the BuddyStudy, we tested the feasibility of using dog fostering as a proxy for dog acquisition, which would allow ethical random assignment. In this single-arm trial, 11 participants fostered a rescue dog for six weeks. Physical activity and psychosocial data were collected at baseline, 6, and 12 weeks. At 6 weeks, mean change in steps/day was 1192.1 ± 2457.8. Mean changes on the Center for Epidemiologic Studies Depression Scale and the Perceived Stress Scale were −4.9 ± 8.7 and −0.8 ± 5.5, respectively. More than half of participants (55%) reported meeting someone new in their neighborhood because of their foster dog. Eight participants (73%) adopted their foster dog after the 6-week foster period; some maintained improvements in physical activity and well-being at 12 weeks. Given the demonstrated feasibility and preliminary findings of the BuddyStudy, a randomized trial of immediate versus delayed dog fostering is warranted.

**Keywords:** dog ownership; dog walking; physical activity; accelerometry; psychosocial well-being; prospective trial; animal-assisted intervention; dog rescue; foster dog; shelter dog

#### **1. Introduction**

Regular physical activity (PA) reduces the risk of cardiovascular disease, type II diabetes, depression, dementia, and some cancers [1]. For individuals living with one or more chronic conditions, regular PA is key to limiting disease progression, preventing co-morbid conditions, and improving physical function and quality of life [1]. The 2018 Physical Activity Guidelines for Americans recommend adults engage in 150 weekly minutes of moderate-intensity aerobic PA (equivalent in intensity to a brisk walk) to reap these health benefits [2]. The most recent national statistics suggest that fewer than one in two American adults meet this mark [3].

A growing international literature base has examined the relationship between dog ownership and PA levels. As discussed in two recent meta-analyses, several studies have demonstrated that dog owners, on average, are more active than non-dog owners [4], and that dog owners are more likely than non-dog owners to meet PA guidelines [5]. Importantly, multiple studies have examined this relationship in clinical populations, including patients with ischemic heart disease [6] and diabetes [7]. The primary limitation of the literature base is its cross-sectional nature; to date, only two studies have prospectively examined the relationship between dog ownership and PA [8,9]. While it is possible that dog acquisition leads to an increase in PA, it is also possible that more active individuals opt to become dog owners. A recent study reporting that dog owners are more likely to be home owners and have a higher annual household income than non-dog owners [10] may support the latter hypothesis, as socioeconomic status is a consistent correlate of leisure-time PA levels [11].

Determining the direction of the dog ownership–human health relationship has important implications. For example, if dog acquisition leads individuals to adopt a more active lifestyle, programming and policies that aim to improve public health might support responsible dog ownership (e.g., encourage pet-friendly lease agreements, provide financial support for veterinary care in low-income communities). Dog ownership could also become a prescriptive tool for physicians to facilitate patient PA, assuming risks are properly considered and mitigated [12]. If renters that can now have a dog in their home or patients that are 'prescribed' a dog choose to acquire a rescue dog, then these initiatives could simultaneously improve human health and dog welfare.

Rigorous prospective studies of the relationship between dog acquisition and PA are needed to inform public health policy and clinical practice. Studies must collect PA data on new owners before and after they acquire dogs, and compare it to data from a group of people that do not acquire dogs. Ideally, a randomized controlled study design would be used to ensure there are no baseline differences between groups that may affect the outcome. Although quasi-experimental (non-randomized) designs and sophisticated analyses [13–16] can help control for known confounding variables (variables that influence both acquisition of a dog and PA outcomes), there may be unknown confounding variables. To our knowledge, only one study used a randomized design to examine how taking a pet into the home affects pet owner health. In 2001, Allen et al. randomized 48 hypertensive individuals to a pet ownership plus ACE inhibitor condition or an ACE inhibitor only condition [17]. Individuals in the pet ownership group acquired a pet cat or dog at the time drug therapy began. Mental stress tests were conducted in participants' homes at baseline and 6 months. The researchers concluded that ACE inhibitor therapy alone reduced resting blood pressure, but that social support through pet ownership reduced the psychological response to mental stress. This study did not assess changes in PA.

Given the serious commitment involved in acquiring a pet, random assignment is no longer considered ethical. The purpose of the current, single-arm trial was to test the feasibility of using dog fostering as a proxy for dog acquisition, as dog fostering is a non-permanent commitment that allows for ethical random assignment. If deemed feasible, a randomized trial of immediate versus delayed dog fostering (or some attention control) would allow for rigorous examination of short-term changes in PA that occur upon taking a dog into one's home. If participants are given the option to permanently adopt their foster dogs after the study period, this approach may also allow for examination of long-term changes in PA. In the BuddyStudy, participants fostered a dog for six weeks and we collected PA, sedentary behavior, and psychosocial outcome data at three time-points to test the feasibility of assessment procedures and examine preliminary effectiveness.

#### **2. Materials and Methods**

#### *2.1. Sample*

BuddyStudy participants were non-dog owners who were willing and able to be the primary caregiver for a rescue dog for 6–8 weeks. Exclusion criteria included (1) <21 years old, (2) self-reported regular exercise over the past 6 months, (3) lack of reliable source of transportation, (4) presence of any conditions that limit ability to walk, (5) presence of uncontrolled hypertension or diabetes, (6) extensive upcoming travel plans, and (7) dog allergy. Inclusion/exclusion criteria aimed to identify a sample of inactive adults who could safely walk for exercise, and whose living arrangements allowed fostering. All participants were recruited through social media outlets affiliated with the University.

#### *2.2. Community Partner*

Dogs involved in this study were fostered through Last Hope K9 Rescue (LHK9), an all-volunteer, all-breed dog rescue organization based in Boston, MA. LHK9 is a foster-based rescue, meaning they rely on foster homes for their dogs and do not have a brick-and-mortar facility. All LHK9 dogs are evaluated and vetted prior to transport from their southern partners in Arkansas, and again prior to entering foster care in New England. Traditionally, LHK9 foster families foster a dog until the dog is adopted, usually between 3–4 weeks, and they are not allowed to adopt their foster dog if there is prior interest from other adopters. Procedures were modified for BuddyStudy participants, who fostered for an extended period of time (minimum six weeks) and were given the option to "foster-to-adopt" as part of the study.

#### *2.3. Study Design*

The BuddyStudy was a single-arm feasibility study. After fostering for six weeks, each participant could adopt the dog, transfer the dog to another foster family, or continue fostering until his/her foster dog was adopted into a permanent home. All participants provided written informed consent to participate in the study and signed a contract, which included a liability waiver, to foster with LHK9. This study was approved by the University's IRB and IACUC.

#### *2.4. Procedures*

#### 2.4.1. Screening Procedures

Initial eligibility was determined via an online screening survey (Qualtrics). Preliminarily eligible individuals attended an orientation at the University where study procedures were explained and written informed consent obtained. Individuals were asked to complete an application to foster via the LHK9 website within 48 hours of attending orientation. LHK9 volunteers followed standard screening procedures for prospective foster home applications, including conducting home visits and calling personal references, current or recent veterinarians, and landlords (when applicable). Individuals deemed ineligible to foster were compensated \$25.

#### 2.4.2. Foster Procedures

Approved participants attended a virtual new foster orientation, and then were added to a private online group moderated by LHK9 volunteers. Each week, a foster coordinator would post a list of dogs (including photos, age, breed and any known background information) needing a foster home in New England. Young puppies (less than one year old), dogs with known medical or behavioral issues, or dogs who already had interested adopters waiting were not eligible for the BuddyStudy. After a participant matched with a foster dog, they picked up an 8-week supply of dog food, a slip

lead, a six-foot non-retractable leash, dog toys, and a crate at the University. Each LHK9 foster dog was microchipped and fitted with a no-slip martingale collar with identification tag.

Per Massachusetts state law, all dogs coming into Massachusetts from out-of-state must be isolated in an approved facility for a 48-hour quarantine period before going to a foster home. LHK9 volunteers told participants what day their foster dogs were ready to be picked up at the quarantine facility, typically with 72 hours notice. If a participant was unable to pick up the dog on the pick-up date, they were asked to notify the research team for coverage. Each participant's 6-week study foster period began the day he/she picked up his/her foster dog.

In addition to the information provided during the virtual orientation, LHK9 provided all BuddyStudy participants with a Foster Information Packet with general dog care information, rescue protocols, and an extensive list of contacts and resources, including a 24/7 emergency line. Participants were asked to post all non-emergency questions in the private online group where a team of volunteers, including LHK9's medical and training coordinators, could respond and provide support. All preventative care (i.e., flea/tick, heartworm) for the duration of the foster period and any medical issues requiring veterinary care were coordinated and paid for by the rescue. Leash and collar safety were discussed, but no specific instructions regarding dog walking were provided.

#### *2.5. Measures*

Data were collected at baseline (pre-foster period), six-weeks (last week of foster period), and 12-weeks (post-foster period). Of note, some participants (adopters) had a dog in their home at 12 weeks and others (non-adopters) did not.

#### 2.5.1. Feasibility

The primary purpose of the BuddyStudy was to determine the feasibility of using dog fostering as a proxy for dog ownership. Formal feasibility assessments focused on recruitment potential (number of completed applications, proportion of applicants deemed eligible, proportion enrolled), participant attrition (proportion dropping out prior to foster and proportion dropping out after getting the dog), data completeness, significant adverse events (number, type), and percentage of dogs adopted.

#### 2.5.2. Device-Measured PA and Sedentary Behavior

The ActiGraph wGT3X-BT monitor (ActiGraph, Pensacola, FL, USA) was used to assess PA and sedentary behavior. The ActiGraph is a research-grade triaxial accelerometer deemed valid and reliable in free-living conditions [18,19]. Participants wore the device on an elastic band at their right hip during all waking hours (except when showering/swimming) for Seven consecutive days at all three time points. During each seven-day assessment period, participants logged all leisure-time PA, including dog walking specifically, to provide contextual information about their activity (as accelerometers only provide data on amount of PA, not type). ActiGraph data were processed using Actilife Version 6.13.3. (ActiGraph, Pensacola, FL, USA) to determine steps/day and PA minutes/day spent in each intensity category (sedentary, light, moderate, moderate-to-vigorous [MVPA]) based on the Freedson cut points [20]. A minimum of three weekdays and 1 weekend day with at least 10 hours of wear time was required for inclusion in analyses.

#### 2.5.3. Self-Reported Dog Walking

Three questions from the Dogs And WalkinG Survey (DAWGS) were used to assess self-report dog walking behavior at six weeks among all participants and 12 weeks among adopters [21]. Questions included, "how many days do you walk your foster dog in a typical week?" (0–7 days); "how much time do you spend walking during your typical dog walk? (minutes); "on days you walk your dog, on average how many walks do you go on?" (1–5 or more). If participants reported a range for dog walking duration (e.g., 10–20 min), the lower value was used to calculate the average dog walking minutes/week.

#### 2.5.4. Psychosocial Outcomes

Stress and depressive symptoms were assessed at all three time points. The 10-item Perceived Stress Scale (PSS) [22], which asks about thoughts and feelings during the last month, was used to evaluate changes in stress. This scale is widely used in behavioral health research and its psychometric properties have been established [23]. Scores can range from 0 to 40 with higher scores indicating more perceived stress. The PSS scale has no standard cut points; as a reference, an average score of 15.5 ± 7.4 was found in a large US sample (n = 968) in 2009 [24]. The 20-item Center for Epidemiologic Studies Depression Scale (CES-D) [25], which asks about feelings and behaviors over the past week, was used to measure changes in depressive symptoms; scores can range from 0 to 60 with scores ≥16 indicating risk for clinical depression. The CES-D has demonstrated high internal consistency, acceptable test–retest stability, concurrent validity, and construct validity when used in general American populations [25]. At 6 weeks, questions used in work by Wood and colleagues [26] were used to assess whether participants got to know people in their neighborhood since starting the study and, if so, whether the foster dog facilitated the interaction and whether the interaction developed into a friendship and/or provided a new source of social support (emotional support, informational support, appraisal support, or instrumental support). Participants also answered open-ended questions about the best part of fostering, the most challenging part of fostering, and the effect of fostering on quality of life.

#### *2.6. Statistical Analyses*

The purpose of this study was to test feasibility and therefore we did not perform inferential statistics. Quantitative data were summarized using means and standard deviations. Qualitative data from open-ended survey questions were coded by two coders (B.M. and C R.) to identify common themes seen throughout the responses. Each coder independently performed a content analysis to identify themes and categories to organize and refine the data. Following individual analyses, the coders compared their results and resolved any discrepancies prior to making conclusions. Direct quotes (de-identified) were extracted from the data to represent the general themes. Analyses of 12-week outcomes were restricted to participants who adopted their foster dog.

#### **3. Results**

#### *3.1. Feasibility*

One hundred and twenty-three individuals applied to participate in the BuddyStudy over the course of 6.5 weeks. About one-quarter of applicants (n = 28; 23%) were deemed preliminarily eligible and invited for study orientation. Eighteen individuals (15% of applicants) completed informed consent to enroll in the study. Twelve received the intervention (i.e., fostered a dog) and eleven completed the study. See Figure 1 for a detailed study flow chart. One participant dropped out of the study within days of picking up her dog, and the dog was transferred to a participant that had not yet matched with a dog. All other participants completed the 6-week foster period and completed 6-week and 12-week assessments. At six weeks, 11 of 11 participants provided valid ActiGraph data; at 12 weeks, 8 of 11 participants provided valid data. There were no significant adverse events. Of 11 rescue dogs in the study, 8 were adopted by their study foster family at the completion of the 6-week study foster period.

**Figure 1.** BuddyStudy flow chart.

#### *3.2. Participant Characteristics*

All participants who received the intervention (n = 12) were female (100%) and the majority were non-Hispanic white (83%). The average age was 37.8 ± 16.3 years (range 21–62 years). All had a college degree (67%) or were current college students (33%). The majority of participants reported living in a rural (42%) or suburban setting (50%); of seven participants with a yard, only one had a fenced area for their foster dog. Participants averaged 6932.7 ± 2418.9 steps/day, 33.6 ± 19.7 MVPA minutes/day, and 572.4 ± 65.3 sedentary minutes/day at baseline. The average score on the stress measure (PSS) was 15.0 ± 6.9 and the average score on the depressive symptom measure (CES-D) was 13.9 ± 12.5 at baseline.

#### *3.3. PA and Sedentary Behavior*

Average steps/day and MVPA minutes/day as measured by the ActiGraph, as well as self-reported dog walking data, are presented by time point in Table 1. At 6 weeks, participants reported dog walking 6.5 ± 0.9 days/week (range 5–7) and increased steps/day by 1192.1 ± 2457.8 and MVPA minutes/day by 12.7 ± 20.9 from baseline. The majority reported 10–15 min walk durations (n = 4) or walks ≥30 min in duration (n = 5). Most participants (n = 10) reported taking more than 1 walk/day on days that they walked their foster dog. Nearly half of the sample (n = 5; 45%) increased their steps by >2000 steps/day and their MVPA by >20 min/day. Individual changes in PA are presented in Figure 2.


**Table 1.** BuddyStudy physical activity (PA) and psychosocial outcomes by time point.

Results reported as mean ± standard deviation unless otherwise noted. <sup>1</sup> Only includes baseline data from participants who completed the study. <sup>2</sup> Adopters, only. <sup>3</sup> n = 6 (n = 2 adopters had invalid ActiGraph data at 12 weeks). Abbreviations: MVPA = moderate-to-vigorous physical activity; CES-D = Center for Epidemiologic Studies-Depression scale; PSS = Perceived Stress Scale.

At 12 weeks, participants who still had a dog in their home (n = 8) reported walking 6.8 ± 0.5 days/week (range 6–7). The majority reported that typical dog walks were ≥30 min in duration (n = 3) or that walks were sometimes as short as 5 min (n = 4). Most still reported walking more than once per day (n = 6) on days that they walked their dog. Adopters who provided valid ActiGraph data at this time point (n = 6) maintained an increase of 552.7 ± 3557.1 steps/day and 8.8 ± 18.8 MVPA minutes/day from baseline.

**Figure 2.** *Cont.*

**Figure 2.** Individual participant changes from baseline to 6 weeks in physical activity, sedentary behavior, and psychosocial outcomes in the BuddyStudy (n = 11). Black bars indicate positive change is better; red bars indicate negative change is better. Physical activity and sedentary behavior were measured via ActiGraph accelerometer. Abbreviations: MVPA = moderate-to-vigorous physical activity; PSS = Perceived Stress Scale; CES-D = Center for Epidemiologic Studies- Depression scale.

Sedentary behavior data are also presented in Table 1. At 6 weeks, participants decreased sedentary minutes/day by 49.8 ± 41.1. Individual changes in sedentary behavior at 6 weeks are presented in Figure 2. At 12 weeks, participants who still had a dog in the home and provided valid ActiGraph data (n = 6) were sedentary for 56.4 ± 59.6 fewer minutes/day, on average, than baseline.

#### *3.4. Psychosocial Outcomes*

Average scores on the stress (PSS) and depressive symptom (CES-D) measures by time point are also presented in Table 1. At 6 weeks, participants reported a reduction of 0.8 ± 5.5 on the PSS and 4.9 ± 8.7 on the CES-D. Additionally, more than half of participants (6/11; 55%) reported meeting someone new in their neighborhood because of their foster dog. While only 1 participant considered someone they met through their foster dog to be a friend, 5 of 11 reported receiving some form of social support through an acquaintance they met through their dog. Individual changes in perceived stress and depressive symptoms are presented in Figure 2. At 12 weeks, participants who still had a dog in their home (n = 8) scored 4.8 ± 7.8 points lower on the stress measure (PSS) and 7.2 ± 12.1 points lower on the depressive symptom measure (CES-D), on average, than at baseline.

Responses to open-ended survey questions at 6 weeks are summarized in Table 2. When asked to describe the best part of fostering a dog, common themes included feelings of fun, happiness, joy, companionship, and family bonding. The most commonly reported challenge of fostering was the stress and responsibility involved with taking care of a dog. Participants reported that fostering a dog either improved their quality of life or that the impact was mixed; no participants reported that fostering had a predominately negative impact on their quality of life.



1 Only themes endorsed by at least three participants are included in the table.

#### **4. Discussion**

Randomized controlled trials provide the highest level of evidence of treatment effectiveness in clinical research. The purpose of the BuddyStudy was to pilot a novel approach that would allow researchers to investigate the PA and health benefits of 'getting' a dog using a randomized controlled design. The BuddyStudy used dog fostering as a proxy for dog acquisition, as fostering is a non-permanent commitment and thereby allows for ethical random assignment. The pilot demonstrated that the design is feasible for implementation on a larger scale, based on the high degree of community interest in the project, high retention among participants who fostered a dog, and absence of significant adverse events. Further, most participants permanently adopted their foster dog, which may allow for the examination of long-term changes in PA with dog acquisition.

The BuddyStudy pilot also demonstrated the preliminary effectiveness of taking a dog into one's home for increasing PA. At the end of the six-week foster period, around half of participants increased their steps by >2000 steps/day and MVPA by >20 minutes/day. These are clinically meaningful increases, as previous research has demonstrated that increases in the range of 1000–2000 steps/day have been associated with reduced risk of type II diabetes [27], cardiovascular disease [28,29], and all-cause mortality [30–32] and the 2018 Physical Activity Guidelines for Americans [2] recommend adults accumulate 150 MVPA minutes/week to improve health and prevent chronic disease. Based on this BuddyStudy data, we anticipate that a sample size of 136 (68/arm) would be needed to detect, with at least 80% power, a difference of 1192 steps/day (change from baseline) between randomized arms in a full-scale trial. Future studies may wish to match self-reported dog walking data with accelerometer data by timestamp [33] to determine the number of steps and MVPA minutes accumulated specifically during dog walking. It should be noted that some participants decreased their daily steps and MVPA minutes from baseline to six weeks. This may have been due to the onset of cold weather, as baseline assessments were completed in September/October and six-week assessments were completed in December in New England.

Many BuddyStudy participants also decreased sedentary time by >45 min/day. There is growing scientific interest in the health risks of too much sitting [34], and some countries now include sedentary behavior guidelines along with PA guidelines. For example, Australia's government recommends that adults minimize the amount of time spent in prolonged sitting and break up long periods of sitting as often as possible [35]. Our findings are in line with three recent studies that have examined dog ownership in relation to sedentary behavior. In a large epidemiological investigation, Garcia et al. found that dog ownership was associated with a lower likelihood of being sedentary for ≥8 hours/day among postmenopausal women [36]. In two separate studies with older adults, dog ownership was associated with an average of 21 fewer minutes of sedentary time/day as measured by ActiGraph accelerometers [37] and with fewer sitting events as measured by active PAL monitors [38].

The BuddyStudy pilot also demonstrated the preliminary effectiveness of taking a dog into one's home for improving psychosocial well-being. At the end of the six-week foster period, most participants who fostered a dog reported decreases in depressive symptoms. When asked to reflect on how fostering a dog affected their quality of life, multiple participants mentioned increased joy, fun, and companionship, which may explain improvements in mood. Alternatively, few participants reported reductions in stress at six weeks. This likely reflects the significant responsibility and time commitment involved in fostering a rescue dog, which many participants acknowledged when asked about the most challenging thing about the foster experience. Finally, many participants reported meeting someone new in their neighborhood because of their foster dog. The phenomenon of dogs as social facilitators has been previously demonstrated [26], and may be one of the most important health benefits of dog ownership given the powerful influence of social relationships, or lack thereof, on human health and longevity [39,40].

To date, no randomized trials have examined the influence of dog 'acquisition' on dog owner PA. Two non-randomized trials have collected self-report PA data on individuals before and after they acquired a dog, and compared it to data from individuals who did not acquire a dog; both found

increases in self-reported recreational walking among new dog owners [8,9]. A large randomized trial of immediate versus delayed fostering that employs device-based PA assessment would build on these studies in two key ways: by better controlling for confounding variables that might explain observed increases in PA (with randomization) and by improving the precision of PA assessment (with accelerometry and a larger sample). A larger sample would also allow for the examination of previously demonstrated correlates of dog walking (e.g., owner income, sense of obligation to walk the dog [41]) as moderators and mediators of change in PA.

A number of lessons were learned from the BuddyStudy pilot and should be considered prior to scale-up. First, given the extensive foster screening process and significant commitment that fostering requires, researchers should anticipate substantial attrition between the informed consent and dog matching steps. Second, running the trial in small waves (n = 5–8) may help to not overwhelm the rescue organization and also account for seasonality, which is likely to impact dog-related PA. Third, future trials should be scheduled to avoid major holidays. In the BuddyStudy pilot, we had to board a foster dog for a week after Thanksgiving coverage plans fell through last minute. Finally, and more generally, the pilot taught us that, while feasible, this study is logistically challenging and will require a sizable research team and budget to be properly conducted. In addition to traditional costs of conducting a randomized trial, the budget should cover costs relating to dog transportation, quarantine boarding, preventive medical care for all study dogs, dog foster supplies (e.g., crates, food, leashes), and reimbursement for short-term boarding, training, and medical care, like dewormer medication and vet visits (all of which are typically covered by rescue organizations).

Finally, while the BuddyStudy study design ultimately uses dog fostering to examine how 'acquiring' a dog influences PA and health, it simultaneously raises awareness about the need for foster homes to reduce shelter euthanasia rates and recruits new volunteers for the cause. An estimated 6–8 million cats and dogs enter US animal shelters each year and 3 million are euthanized [42]. Across the US, shelters are filled with dogs that need to be cared for and exercised by volunteers. Foster families are also needed to make room in the shelters for more surrendered and abandoned dogs. Foster-based dog rescue organizations, like LHK9 Rescue, transport dogs from shelters in overpopulated parts of the country to areas that tend to have fewer dogs for people to adopt locally. The more foster families these organizations have, the more dogs they can pull in to rescue. All 11 rescue dogs involved in the BuddyStudy pilot were permanently adopted, eight by their study foster parents.

The primary limitations of this study are the small sample size and lack of control group, which limit our ability to draw conclusions about effectiveness. The study sample was also highly homogenous (100% women; 83% non-Hispanic white). Future studies, including multi-site trials, could examine effectiveness in a more diverse population and across different geographic regions. It should be noted that the vast majority of animal welfare volunteers are women, and therefore this approach may ultimately appeal more to women [43]. Importantly, women are less active and therefore in greater need of PA intervention than men [3]. Finally, our two primary psychosocial variables were assessed over different time frames (past week for depressive symptoms, past month for perceived stress) and future studies should consider using measures with more similar time frames. Strengths of this study include the innovative approach taken to address a critical gap in the literature, which required a novel University-Dog Rescue partnership. The use of accelerometry to measure PA is also a strength, as the literature on dogs and PA is heavily reliant on self-report PA measures prone to recall and social desirability biases.

#### **5. Conclusions**

Given the demonstrated feasibility and preliminary findings of the BuddyStudy pilot, a randomized trial of immediate versus delayed dog fostering is warranted and will provide the most rigorous evidence to date of the effects of dog acquisition on human PA and health. Given the major disease and economic burden caused by physical inactivity [44], as well as the ongoing pet homelessness epidemic [42], this line of research has significant societal implications. Demonstration of a positive causal relationship between dog acquisition and increased PA could lead to the testing and implementation of evidence-based public health policies and programs that encourage responsible dog ownership.

**Author Contributions:** Conceptualization, K.P.; methodology, K.P.; validation, K.P.; formal analysis, K.P., B.M. and C.R.; investigation, K.P., B.M., C.R.; resources, K.P. and J.E.T.; data curation, K.P.; writing—original draft preparation, K.P.; writing—review and editing, J.E.T., B.M., C.R., and L.B.B.; visualization, K.P.; supervision, K.P. and J.E.T.; project administration, K.P. and J.E.T.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors thank Last Hope K9 Rescue and all of its volunteer coordinators and directors for the significant time, energy, and resources dedicated to this research project and to responsible rescue; we want to also thank LHK9 foster families in both Arkansas and New England who give their homes and hearts to Last Hope rescue dogs. The authors thank Nick Diehl and Grace McCarthy for their assistance with foster screening procedures. Potter thanks Rena Wing for her support and assistance in conceptualizing the study design.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Investigating the Role of Prolactin as a Potential Biomarkerof Stress in Castrated Male Domestic Dogs**

**Jara Gutiérrez 1,\*, Angelo Gazzano 1, Federica Pirrone 2, Claudio Sighieri <sup>1</sup> and Chiara Mariti <sup>1</sup>**


Received: 2 August 2019; Accepted: 10 September 2019; Published: 12 September 2019

**Simple Summary:** Although cortisol is usually considered the main reference for the assessment of stress, in some animal species it has been shown that prolactin can be used as a biomarker of both acute and chronic stress. Behavioural parameters can also be used to assess the state of welfare and stress. This study was aimed at evaluating the possible relationship between serum prolactin, serum cortisol and behavioural signs of stress in domestic dogs. To reduce the possible influence of some factors, the study was performed on a homogeneous sample formed by 40 castrated male Spanish Greyhound dogs housed in a dog shelter. The weak negative correlation found between serum cortisol and prolactin values agrees with results obtained in other studies, indicating that prolactin response might be an alternative to cortisol response.

**Abstract:** Prolactin has been recently regarded as a potential biomarker of both acute and chronic stress in several species. Since only few studies until now have focussed on domestic dogs, this study was aimed at evaluating whether prolactin, cortisol and stress behaviour correlated with each other in sheltered dogs. Both cortisol and prolactin analysis were performed in serum samples through a hormone-specific ELISA kit. For each dog, a stress score was calculated by summing the number of occurrences of stress-related behaviours. The presence/absence of fear during the time spent in the collection room was also scored for each individual. Results revealed a weak negative correlation between cortisol and prolactin levels. Neither of the hormones was correlated with the stress score, nor did their values seem to be influenced by showing fear in the collection room. The weak negative correlation found between cortisol and prolactin values agrees with results obtained in other studies, indicating that prolactin response might be an alternative to cortisol response. This, together with the high serum prolactin levels compared to those reported by other authors for healthy domestic dogs, may indicate that prolactin might be a good biomarker of chronic stress, and although further studies are needed to better understand the potential role of prolactin in the evaluation of canine welfare.

**Keywords:** behaviour; cortisol; dog; male; prolactin; shelter; Spanish Greyhound; stress

#### **1. Introduction**

Although prolactin hormone is known for its function in the stimulation of the growth of the mammary gland and the lactation processes, it has more than 300 different biological activities, homeostatic roles and physiological functions in the organism, e.g., electrolyte balance, luteal function, regulation of the immune system, osmoregulation, angiogenesis, and maintenance of the inter-oestrous interval [1]. In fact, the important genetic role of the prolactin receptor (PRL-R) in energy balance and metabolic adaptation has been recently evidenced in rodents, proving that prolactin has essential roles for the metabolism of glucose, insulin and lipids, as well as in promoting a positive energy balance [2]. Furthermore, prolactin has been found to be an index of acute stress in some non-human animal species, such as rats [3–5], domestic ruminants [6], donkeys [7], cattle [8], and sheep [9]; and different

specific functions of prolactin during the stress response have also been demonstrated [2], such as mediation in the epidermal adaptation to environmental stress in fish [10].

It has been recently found that prolactin signalling in tuberoinfundibular dopamine (TIDA) neurons was reduced in mice exposed to acute stress, with a consequent potential decline in their inhibitory influence on prolactin secretion [11]. This suggests that prolactin secreted in response to acute stress may activate prolactin receptors in certain tissues involved in physiological adaptation to stress [11]. In humans, prolactin is stimulated by suckling, perception of visual, acoustic and olfactory stimuli, as well as by stress [12], including psychological and psychosocial stresses [13], and experimental stress-related conditions, such as hypoglycaemia [14], surgery [14], parachute jumping in military recruits [15], and compulsory swimming in non-swimmers [16]. In fact, cortisol, prolactin and day-to-day changes in anxiety indexes measured by questionnaires were found to be significantly correlated [17].

However, the scientific literature reports divergent results with both unchanged [18,19] and decreased prolactin levels [20–22] in response to stress, questioning the belief that stress stimulates prolactin secretion [23]. In addition, while cortisol levels were found to correlate with anxiety levels the day before a surgical intervention, no significant correlation between prolactin and anxiety was found [24].

As for the domestic dog (*Canis familiaris*), to our knowledge the potential link between stress and prolactin levels has been scarcely investigated. One study assessing surgery-related stress in dogs found that cortisol increased and prolactin decreased in the post-surgery period compared to basal levels [25], the latter being in contrast with increased post-operative prolactin response found in humans after surgery [26]. The authors suggested that this contrast may be showing that different species can exert differences in the activation of prolactin feedback regulatory systems [25].

A few studies have reported that dogs with generalized anxiety had hyperprolactinemia, but dogs with phobias or mild anxiety did not [27,28]. Prolactin blood levels were positively correlated with the score obtained on an individual scale for quantifying the presence of anxiety-related behaviours through a physical and behavioural evaluation [28]. In fact, dogs suffering from different kinds of generalized anxiety showed a significant decrease in dopamine blood levels [29], which can justify measuring prolactin, since its secretion is mainly controlled by dopaminergic neurones [28]. However, the involvement of prolactin in emotional responses in dogs seems to be more complex, as its circulating concentrations have been shown to increase also during positive interaction with humans [30].

Cortisol is commonly regarded as the "stress hormone"; cortisol and its metabolites have in fact been quantified in various sample matrices such as blood, saliva, urine and faeces [31], and more recently in hair [32–34], as an endocrine response to stress. Cortisol levels have been shown to increase in adverse conditions such as isolation, restriction of movement, regrouping or transport [35]. In fact, cortisol is secreted following the activation of the hypothalamic–pituitary–adrenal (HPA) axis, one of the major stress response systems [36]. Nonetheless, increased HPA activity is not stress-specific, since it can also be caused by metabolic processes, positive affective states, maternal behaviour and physical activity [37–40]. For this reason, some authors suggest that an increase in cortisol levels should be regarded as an indicator of arousal [37] rather than of stress.

The reliability of cortisol in the assessment of stress has further been questioned due to many factors that can affect the interpretation of cortisol levels such as individual variability in the response to a stressful exposure [41,42], a high inter-individual variability in baseline cortisol levels between dogs [43], distress for blood sampling [44], and possible variations due to the circadian rhythm [44], the latter being not clearly identified in domestic dogs: in fact, no circadian rhythm for cortisol has been reported for laboratory [45] and working dogs [46,47] for intervals of 24–28 h. All these factors suggest caution when comparing cortisol levels in groups of different individual dogs, and when assessing long-term cortisol secretion from fluids such as blood, urine and saliva [44]. The evaluation of cortisol levels can be further complicated by changes with respect to the duration of stress. Whereas in the presence of an acute stressor serum, cortisol rapidly increases and then returns to the basal levels, in a chronic stress situation, a prolonged exposure to the stressor can lead to the suppression and deregulation of the HPA axis [48], which generally does not mean a normalization of circulating cortisol levels, and thus, they are a very useful measure of chronic stress [49].In recent years, hair cortisol analysis has also been supported as a reliable reflection of long-term cortisol secretion [50,51].

Cortisol and prolactin seem to both be involved in the stress response and their levels are likely to be somehow associated. For instance, in vitro experiments showed higher basal levels of corticosterone in hyperprolactinemic rats than in normal and hypoprolactinemic ones; and prolactin was found to exert a stimulating effect on ACTH-induced corticosterone secretion in acute restraint stress [52]. Nevertheless, the response of both prolactin and cortisol hormones to a stressful situation may be divergent, and some authors suggest that prolactin release can act as an alternative form of the cortisol response to stress [22].

In dogs, the evaluation of stress is often a combination of physiological and behavioural parameters, the latter being usually considered a reliable indicator [53,54]. For instance, after the induction of chronic stress through a model of social and spatial restriction, dogs living in small indoor kennels showed significantly lower postures [55], as well as an increase in cortisol levels [56] compared to those living in enriched spacious outdoors housing in groups. When exposed to short-kennelling environment, dogs were generally more active [36,57], and their cortisol:creatinine ratios (C/Cr) were significantly higher compared to when they were in home environment [36].

The aim of this opportunistic study was to evaluate whether there is a correlation between serum cortisol and prolactin concentrations and between them and behavioural indicators of stress and fear in domestic dogs. To reduce the impact of possible affecting factors such as sex, neutering state, breed and housing condition, this study was carried out on a homogeneous sample of castrated male Spanish Greyhound dogs housed for more than 155 days in a dog shelter. This breed has a complex social context in the south of Spain, where this study was performed, with reports of abuse and neglect being frequent for them. Therefore, they could have experienced previous common adverse experiences prior to being rescued, which could reinforce the homogeneity of the sample.

The hypothesis of this study was that, as previously suggested by other authors, prolactin could be used as a biomarker of stress in domestic dogs and, for this reason, that it was correlated with other measures related to stress, such as serum cortisol, the presence of fear or the frequency of stress-related behaviours in dogs. Prolactin secretion from the pituitary gland is used as a marker for lactotropic axis activation [58], it has been found to be an index of acute stress in some species, and have specific functions during the stress response [2]. However, due to the controversial findings in the literature, we did not make any predictions about the possible increase or decrease of prolactin levels in a stressful condition.

This study includes some elements of novelty such as being, to our knowledge, the first including multiple measures of stress, both hormonal and behavioural (prolactin, cortisol and behaviours related to stress and fear), investigating their possible correlation, and involving such a homogeneous sample of domestic dogs, with all belonging to one single breed, being castrated males, and living in the same environment.

#### **2. Materials and Methods**

The procedure was communicated to the Ethics Committee of the University of Pisa, Italy and it received a favourable opinion with Decision N.09/2018.

#### *2.1. Subjects and Place*

Samples were collected between 3rd September and 26th November 2018 from 40 sheltered castrated adult male Spanish Greyhound dogs (mean age ± standard deviation = 46.5 ±20.8 months; min. = 19; max. = 112 months). Dogs were housed at the Fundación Benjamin Mehnert (Seville, Spain).

Dogs that entered the shelter were either wandering alone, probably after having been abandoned, occasionally hit by a car, or coming directly from hunter discards or reports of abuse and mistreatment. All of them were hosted in the shelter for a period longer than 155 days (270.9 ± 115.6 days), and had a healthy condition, regularly checked by a veterinarian.

The shelter included a covered space of 2000 m<sup>2</sup> divided into 2 rows and 3 main corridors, with a total of 60 boxes, each of which was provided with 1-3 plastic beds (88 × 58 × 28 cm), depending on the number of dogs sharing the box. Apart from the feeder and the water bucket, there were no other objects such as toys in the boxes. Dogs went out to the outer courtyards in small groups twice a day.

Dogs were usually living in a group of 3–4 dogs (92.5% of cases), all Spanish Greyhounds, sharing the same box for days or weeks; in several cases (n = 3), five dogs were housed together. The shelter staff organized dogs in groups that ensured the greatest possible social stability, thus partially counterbalancing the risk of stress due to overcrowding or conflicts.

#### *2.2. Video Analysis*

A 10-min video was recorded of each dog using a video camera (JVC GZ-MG130E, Yokohama, Japan) located about one meter away from the box fence (the maximum possible distance to cover the whole box) and at a stable height of 100 cm. Videos were recorded between 8:30 a.m. and 3:00 p.m. To facilitate the recognition of each dog in the videos, their most peculiar physical characteristics were registered, and a picture was taken.

Videos were analysed using BORIS (Behavioral Observation Research Interactive Software, University of Turin, Turin, Italy) [59]. A list of 18 possible signs of stress in dogs was created [54,60–63], and for each of them the number of displays (observations) was recorded. A stress score was calculated for each dog, by summing up the total number of times each of the following stress-related behaviour was displayed: yawning, shaking, paw lifting, tongue out, eliminating, growling, turn head, tuck tail, cowering, trembling, circling, pacing, hiding, panting, salivation, howling, whining and behaviour against the fence (any behaviour in which the dog physically contacted the fence, such as jumping on the fence or standing up against it).

Videos were analysed by an observer having experience with video analysis of dog behaviour. To measure observational accuracy, a second observer, also having previous experience in behavioural analysis of videos, analysed 3 videos using the same software (BORIS), and inter-observer agreement was calculated.

#### *2.3. Presence of Fear in the Collection Room*

After recording the video, 2–3 dogs were contemporaneously led on the leash to a room within the shelter facility, from here called collection room. This movement required around two minutes. In the collection room, dogs were restrained using a leash tied to a table.

The presence of fear around the time of blood sampling was qualitatively scored (yes/no), as a result of a global consideration of the dog behaviour during the time spent in the collection room (20.6 ± 7.0 min). Fear was scored with a "yes" for those dogs expressing at least one of the behaviours related to fear [64], such as the tail tucked, trembling, cowering, and hiding behind objects or in a corner, avoidance of eye contact and approach, and freezing.

#### *2.4. Serum Samples Collection and Storing*

Blood (3–4 mL) was drawn from the cephalic vein in the collection room by a veterinarian, part of the shelter staff and familiar to all the dogs, avoiding stress for the dog as much as possible. Only the veterinarian was involved in both restraining the dog during the collection and blood withdrawing.

Based on previous literature about the times at which prolactin and cortisol reach their maximum values and decreased in blood after a stressful situation in humans [65,66], samples were collected about 10 min (ranging from 5 to 15 min) after entering the collection room. When the procedure (identification, blood sampling, and clinical examination) was completed for all the dogs in room, they were moved together to their box.

Tubes containing the blood were kept upright refrigerated (4 ◦C) in a padded box. After 50–90 min, blood was centrifuged at 3000 rpm for 20 min (using Nahita 2615 Auxilab SL, Beriain, Spain). Serum samples were stored in Eppendorf tubes kept vertically in the freezer (−18 ◦C) and maintained frozen until being analysed.

#### *2.5. ELISA Kit*

Prolactin and cortisol concentrations from canine serum were measured using an EIA (enzyme immunoassay) kit (Demeditec Diagnostics®, Kiel, Germany for prolactin; Diametra®, Segrate, Italy for cortisol), according to the manufacturer's instructions.

#### *2.6. Statistical Analysis*

A Shapiro-Wilk test (*p* < 0.05) was applied to the three parameters (prolactin, cortisol, and stress score) to investigate the normality of data. None of them was found to have a normal distribution; therefore, further analyses were carried out using non-parametric statistics.

A comparison between dogs showing fear and dogs not showing fear in the collection room was made for each parameter, through a Mann-Whitney U-test (*p* < 0.05).

A possible correlation between serum prolactin levels (ng/mL), serum cortisol levels (ng/mL) and stress score was investigated using Spearman's rank correlation coefficient (*p* < 0.05).

#### **3. Results**

Regarding the videos, the percentage of inter-observer agreement was excellent (92.1%).

Results of stress-related behaviours are reported in Table 1, showing the number and relative percentage of dogs displaying each of the 18 analysed behaviours, and the number of occurrences (times the behaviour was displayed). Tongue out behaviour occurred with the highest frequency (total frequency = 101) and tuck tail with the lowest (total frequency = 1). Four behaviours (growling, hiding, salivation and trembling) were not displayed for any of the dogs (Table 1).

**Table 1.** Number of dogs and relative percentage (%, calculated on the total sample of 40 dogs) displaying each of the 18 stress-related behaviours, and number of occurrences (times the behaviour was displayed).


Thirteen dogs showed fear in the collection room and 27 did not. Results of the three parameters for individual dogs are reported in Figure 1.

Prolactin values varied between 2.51 and 43.75 ng/mL, with a mean ± standard deviation of 10.82 ± 9.80 ng/mL. No difference in prolactin concentrations was observed when comparing dogs with and without fear in the collection room (*u* = 153.00; *p* = 0.516; see Figure 2).

Cortisol values ranged 4.43–85.14 ng/mL (25.97 ± 20.33 ng/mL). No difference was found in cortisol levels between fearful and non-fearful dogs in the collection room (*u* = 160.00; *p* = 0.654; see Figure 2).

Stress score values varied between 0 and 27 (6.15 ± 6.77). Stress scores did not differ between dogs with and without fear in the collection room (*u* = 143.00; *p* = 0.344; see Figure 2).

Serum prolactin levels and serum cortisol levels showed a weak negative correlation (Spearman Rho = −0.319; *p* = 0.045). The stress score did not show a correlation with either prolactin (Spearman Rho = 0.017; *p* = 0.918) or with cortisol levels (Spearman Rho = −0.155; *p* = 0.342).

**Figure 1.** Results of serum prolactin, serum cortisol and stress score. Values for prolactin in serum (ng/mL), cortisol in serum (ng/mL) and stress score for each dog (n = 40) are shown. Data are organized according to increasing value of prolactin.

**Figure 2.** Results of serum prolactin, cortisol and stress score for dogs showing and not showing fearful in the collection room. Values for prolactin in serum (ng/mL), cortisol in serum (ng/mL) and stress score in dogs evaluated as showing (yes; n = 13) or not showing (no; n = 27) fearful behaviour in the collection room.

#### **4. Discussion**

Scientific literature about prolactin normal ranges in dogs is relatively scarce. It is well documented, considering its role in lactation, that prolactin increases in pregnant and above all in lactating bitches, as well as in pseudopregnant bitches [67,68], with values ranging from 0 to around 40 ng/mL at 10 weeks after oestrus [67,69]. As for healthy male dogs, studies carried out on intact males have found different but similar ranges [70–72], which can be summarised as a normal range in intact male ranging from

non-detectable to 6 ng/mL [71]. Some factors such as breed seem to be responsible for differences in prolactin secretion in intact male dogs [71,72], while age seems not to be influential [71]. The results of the current study refer to castrated males, to dogs housed in a shelter, and to a canine breed not previously involved in studies on prolactin, therefore their interpretation should be performed cautiously. However, data reported in Figure 1, as well as the mean value, show that most of the involved dogs had prolactin concentrations higher than the threshold indicated by Corrada et al. [71], and some dogs had values closer or higher than those reported by Pageat et al. [28] for anxious dogs treated with fluoxetine or selegiline (around 13 ng/mL).

Results on serum cortisol concentrations provided by this study showed a high individual variability, in accordance with Bennett and Hayssen [43]. The values obtained here are included within the wide range (5–100 ng/mL) for baseline cortisol measured by an ELISA kit in client-owned dogs' serum [73]. However, other studies have reported a narrower range, varying according to the physiological state of dogs and the used analytical methods [74–77], making it difficult to compare and frame the values found in the current study. Generally speaking, the mean serum cortisol value obtained in this study was higher than the expected for healthy domestic dogs. The high cortisol values might be due to many factors, including a potential chronic stress condition lived by dogs in the shelter, with a consequent deregulation of HPA axis [48,56]. This will be discussed later.

Looking at the relationships between the investigated parameters, serum prolactin values were not correlated with stress score and they were weakly negatively correlated with serum cortisol levels. Several authors have reported that stressful situations can lead to higher circulating prolactin levels in various species [3–9,11–16], suggesting that prolactin might be a good biomarker of acute stress. Results of the current study seem not to support this hypothesis. However, other studies found that stress-linked cortisol changes did not affect prolactin levels [19,78,79] or have even reported a negative correlation between cortisol and prolactin levels [21,22,30,80,81]. Sobrinho et al. [22] suggested that this negative correlation might be due to the fact that prolactin release acts as an alternative form, rather than an extension of, the more common cortisol response, thus each hormone could be released in response to specific emotions [22].In fact, neural pathways responsible for cortisol and prolactin responses to stress are different in rodents [82], who often release prolactin in response to acute psychological stress [83]. Likewise, in sheltered dogs, whereas serum prolactin levels decreased in response to stress, no changes were observed in serum cortisol concentrations [58].

Prolactin has a direct influence on oxytocin secretion, and both hormones modulate the neuroendocrine acute stress response during maternal behaviour, having an anxiolytic effect in pregnant and lactating rats [84,85]. However, prolactin response to stress does not seem to be limited to females: in a study by Siracusa et al., there was no significant interaction between gender and change in serum prolactin concentrations among time points for dogs exposed to a synthetic, dog-appeasing pheromone both before and after surgery [58].

It must be noticed that the way in which individuals respond to stress and fear may depend on their individual history. Dogs of the current study have experienced situations of abuse, neglect and/or abandonment prior to entering the shelter, which can be all considered stressful situations. It can be hypothesized that their individual history may significantly have affected their way to respond to an acute stress situation. In fact, dopamine release has been shown to increase in response to acute stress in rats if they had previous exposure to chronic stress [86]. This was in agreement with the decrease found in prolactin concentrations in sheltered dogs with assumed chronic stress that were exposed to acute stress [25].

In the current study, prolactin concentrations did not correlate neither to stress score nor to fear behaviour in the collection room. This result, together with the shortage of publications on prolactin response to stress in dogs [58], strongly suggests the need for future research, focused on acute stressors that might provide additional information on the possible use of prolactin as a biomarker of acute stress in domestic dogs. It must be noted that most dogs did not show fearful behaviours in the collection room. This could not be a consequence of their familiarity with this place, since visiting veterinary facilities occurred only for veterinary care, and surgical procedures. However, the presence of both familiar people (shelter staff) and familiar dogs (box mates) likely decreased the stress of novelty and blood withdraw in the collection room.

A possible explanation for cortisol results of the current study is that some dogs belonging to the sample could be chronically stressed and their cortisol levels mirroring such condition: in fact, sheltered dogs are frequently exposed to stressful conditions [87], and those of the current sample had experienced abandonment, abuse and/or neglect.

In other terms, the absence of correlation found between cortisol levels, stress score and fear behaviour in the collection room, as previously reported for the relationship between cortisol and prolactin levels, might be explained by the fact that cortisol values were affected by factors other than those analysed in this study. For instance, the long permanence in the shelter, lived by dogs in our sample, might have led to chronic stress, leading to the suppression and deregulation of the HPA axis [48]. In fact, several studies have reported cortisol variations depending on how long dogs have been in the shelter [88–92]. This point, together with other weaknesses previously reported for cortisol, make even more relevant the need for alternative easily measurable biomarkers of stress.

It could be argued that, if prolactin levels remain constantly elevated during a long period of stress, then this hormone could be a good biomarker of chronic stress in the dog. If prolactin levels in long-stay sheltered dogs were elevated, while cortisol levels varied according to the welfare state of the dog, e.g., decreasing along with the permanence time in the shelter [90–92], then a negative correlation between prolactin and cortisol would be expected, as has been partially reported in the current study. It is advisable that further research investigate whether prolactin is a good candidate for being a biomarker of chronic stress in dogs. For this, it must be taken into account that, at the moment, the use of biological matrixes other than blood for measuring prolactin is not advisable in dogs [93].

This study presents some weaknesses and limitations. One limitation is that this study was focused on assessing the possible association between prolactin and other measures of stress, but no analysis was performed to investigate the possible impact of factors such as the number of dogs per box, the variability on previous individual experiences, etc. In addition, shared experiences such as living in the same cage might have an impact and lead to a cage effect, which was not investigated in the current research. Future studies, involving a wider sample, should take these factors into account. Another limitation is that, although on one hand the homogeneity of the sample may have reduced variability in the data, the specific characteristics of dogs (breed, sex, sexual status, previous experience, and management) require caution in generalizing results to the whole canine population, in particular to pet dogs, whose previous experiences might lead them to respond differently to stress. A potential weakness of this study is related to the lack of knowledge about the timing of prolactin release and decline after a stressful event, especially in the canine species. Future research should better investigate this point.

#### **5. Conclusions**

The study of the serum levels of cortisol and prolactin, and stress behaviour in dogs has not resulted in any positive correlation between variables, and the display of fearful behaviour at the moment of blood sampling seems to not have influenced the assessed parameters. Cortisol levels may depend not only on the observational factors scored in this study (stress score and fearful behaviour), but also on the possible presence of a chronic stress. This might explain the weak negative correlation found with prolactin values. This weak negative correlation, together with the high serum prolactin levels compared to those reported by other authors for healthy domestic dogs, may indicate that prolactin could be a good biomarker of chronic stress.

The evaluation of welfare in kennelled animals is essential to ensure and improve their quality of life, and the importance of a multiparametric approach has been stressed [48]. The study of behavioural and physiological parameters other than cortisol might be very helpful to increase the robustness of these studies and consequently provide a useful tool to facilitate adoptions, apply proper environmental enrichment improvements, or target dogs for behavioural or clinical intervention.

**Author Contributions:** First Author (J.G.) was responsible for data curation, investigation, and writing (original draft + review & editing). Second author (A.G.) was responsible for conceptualization, supervision, and writing (review & editing). Third author (F.P.) was responsible for methodology, interpretation of data and writing (review & editing). Forth author (C.S.) was responsible for writing (review & editing). Last author (C.M.) was responsible for conceptualization, methodology, interpretation of data, supervision, and writing (review & editing).

**Funding:** This research received no external funding.

**Acknowledgments:** We appreciate the collaboration of the dog shelter *Fundación Benjamin Mehnert* (Seville, Spain). **Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*

### **Socioeconomic Influences on Reports of Canine Welfare Concerns to the Royal Society for the Prevention of Cruelty to Animals (RSPCA) in Queensland, Australia**

### **Hao Yu Shih 1,\*, Mandy B. A. Paterson 1,2 and Clive J. C. Phillips <sup>1</sup>**


Received: 28 June 2019; Accepted: 18 September 2019; Published: 23 September 2019

**Simple Summary:** The role of the socioeconomic status of dog owners in canine welfare concerns is not fully understood. We conducted a retrospective study of 107,597 canine welfare complaints attended by the Royal Society for the Prevention of Cruelty to Animals (RSPCA) Queensland from 2008 to 2018. We explored the relationship between the owner's socioeconomic status and reported (rather than confirmed) complaints about welfare of dogs. The socioeconomic status of the owner was estimated from the postcode of where the alleged welfare issue occurred, using government statistics for Socio-Economic Indexes of different regions of Australia. Reported complaints were correlated with socioeconomic scores. There was a lower median socioeconomic score in our study group compared to the entire Queensland state, indicating that alleged canine welfare concerns were more likely to be reported in areas with inhabitants of low socioeconomic status. The status was also low if the complaint was about a crossbred rather than a purebred dog. Among the purebred dogs, complaints involving working dogs, terriers, and utility breeds were associated with the lowest socioeconomic scores. The following complaints were associated with low socioeconomic status: cruelty, insufficient food and/or water, a dog not being exercised, a dog being confined/tethered, failure to provide shelter or treatment, overcrowding, a dog being in poor condition or living in poor conditions. Increased status was observed in alleged cases of a dog being left in a hot car unattended.

**Abstract:** Human–dog relationships are an important contributor to the welfare of dogs, but little is known about the importance of socioeconomic status of the dogs' owners. We conducted a retrospective study of canine welfare complaints, using Australian government statistics on the socioeconomic status of the inhabitants at the location of the alleged welfare issue. The socioeconomic score of inhabitants at the relevant postcode was assumed to be that of the plaintiff. Our dataset included 107,597 complaints that had been received by RSPCA Queensland between July 2008 and June 2018, each with the following information: the number of dogs involved, dog(s) age, breed(s), suburb, postcode, date received, and complaint code(s) (describing the type of complaint). The median index score for relative social advantage of the locations where the alleged welfare concern occurred was less than the median score for the population of Queensland, suggesting that welfare concerns in dogs were more commonly reported in areas with inhabitants of low socioeconomic status. It was also less if the dog being reported was not of a recognised breed, compared to dogs of recognised breeds. Dogs reported to be in the gundog breed group were in the most socioeconomically advantaged postcodes, followed by toy, hound, non-sporting, working dog, terrier, and utility breed groups. Reports of alleged cruelty, insufficient food and/or water, a dog being not exercised or being confined/tethered, failure to provide shelter or treatment, overcrowding, a dog being in poor condition or living in poor conditions were most likely to be made in relation to

dogs in low socioeconomic postcodes. Reports of dogs being left in a hot vehicle unattended were more likely to be made in relation to dogs in high socioeconomic postcodes. It is concluded that both canine welfare complaints and dogs in specific breed groups appear to be related to the owner's socioeconomic status. This study may be used to improve public awareness and to tailor educational campaigns toward different populations.

**Keywords:** dog; animal welfare; animal cruelty; RSPCA; shelter; socioeconomic

#### **1. Introduction**

Dogs (*Canis familiaris*) share intimate relationships with humans as they are one of the most popular pets in our society [1]. According to the 2016 statistics, 3,555,000 Australian households (38.5%) own at least one dog, accounting for 4,759,700 dogs in Australia [1]. In the household, owners may develop a strong attachment to their dogs and vice versa, which creates a mutually beneficial relationship [2,3]. For instance, dog ownership apparently reduces owners' physiological and psychological stresses [3,4], as well as providing a safe haven for their dogs [5]. However, the human–dog relationship does not always provide mutual benefits and sometimes may even break down [6].

Animal cruelty refers to any morally or legally unacceptable behaviour which causes animals to experience physiological, psychological, and/or behavioural discomfort [7,8]. Animal welfare concerns are reported in different forms and in different species, and dogs are one of the most commonly reported victims [9]. Concerns may involve violence [10,11], injuries caused by their involvement in the sporting industry [12,13], abandonment [14,15], inappropriate surgical procedures [16], and neglecting care of the dog [9,17]. Not only do these issues compromise the welfare of dogs, some of these welfare concerns (e.g., intentional abuse and dog fighting) have been recognised as sentinels of other social issues [6,9], in particular domestic violence [18–20] and sexual assaults [21]. Some have also been reported to be a precursor of antisocial behaviours among young people [7,22,23].

To address these cruelty issues, animal protection legislation has been enacted worldwide [8,9,24]. In Queensland, Australia, animals are protected by *Animal Care and Protection Act 2001* (ACPA) [8]. This state-based legislation empowers the State to appoint inspectors, some of whom are employed by the Royal Society for the Prevention of Cruelty to Animals, Queensland (RSPCA Qld), who investigate potential breaches of the Act and enforce compliance with the Act [8]. There are two main offences under the ACPA: failure to fulfil 'duty of care' responsibilities and cruelty. There are a number of other specified offences. The Act recognises that a person who has charge of an animal owes that animal a duty of care. Failure to provide such care is the basis of the "breach of duty of care" offence. This offence covers such actions as not providing sufficient food, water, exercise, veterinary care, and suitable living conditions. It is not only the owner that has a duty of care towards an animal; anyone who is even temporarily in charge of an animal has this responsibility. The second major offence is "animal cruelty" and according to Section 18 of the ACPA cruelty describes any action that causes unjustifiable and unnecessary physical and mental discomfort to animals, inappropriate confinement or transport, unreasonable injuries and inhumane death [8]. A cruel act can be committed by anyone towards an animal, whether it is their own animal, another domestic animal or even a wild animal [8]. It is important to note, that under the ACPA, it is not necessary for a person to have the intention of being cruel for the offence to be proven in Queensland. If an action carried out by a person causes pain and suffering and the action was intentional, the person may be charged with cruelty. The intention to carry out the action must be proved but not the intention to be cruel. If a lack of action deprives an animal of its fundamental needs, then the person who has a duty of care towards the animal may be charged with a breach of their duty of care or cruelty depending on the circumstances. Intention may be considered during sentencing however [8]. Other offences under the Act include unreasonable abandonment or release, the carrying out of prohibited surgical procedures (e.g., tail docking, ear

cropping, debarking, etc.); being involved in, or having items used for, a prohibited event, such as dog or cock fighting; and allowing an animal to injure or kill another animal [8].

Potential cases are reported to RSPCA Qld through various means. RSPCA Qld has a "Cruelty Complaints" telephone number manned 24 h a day, seven days a week and complaints also come in through emails. These complaints are primarily made by members of the public but a few are also made by veterinarians and veterinary nurses, council officers, and other government and non-government employees visiting a location as part of their duties. Animals surrendered to the RSPCA or that come in as strays may be investigated by RSPCA Qld inspectors if cruelty or neglect is suspected. All cases reported by sources mentioned above were the focus of this study.

Apart from aiding law enforcement, identification of risk factors associated with animal welfare concerns would be beneficial in the development of awareness and in education of the general public. Risk factors include dogs' breed or breed group [12,15,25–27], dogs' ages [15,28], behaviours [15,26], owners' characteristics [15], and the complainant's socioeconomic status [6,20,22,29]. Among all these risk factors, the role of household socioeconomic level in crimes and animal welfare concerns is not yet well understood [6,20,22]. It has been proposed that socioeconomic status is negatively correlated with the frequency of crimes, domestic violence, and neglecting and abusing animals [6,22,29–31]. However, a tautologous relationship has also been postulated, because a dog owner's socioeconomic status may affect his or her ability to provide for all aspects of good welfare.

Therefore, this study aimed to investigate the relationships between socioeconomic status, dog breeds, and different types of dog welfare complaints. The socioeconomic status was quantified using a socioeconomic score derived from the postcode of where the alleged welfare issue occurred and government statistics for Socio-Economic Indexes of different postcode regions of Australia. We hypothesized that owners from relatively poor socioeconomic postcode regions would be more likely to be the subject of complaints about an absence of key resources for dogs, such as insufficient food, water, living space or veterinary care, lack of shelter, and poor living conditions. We also hypothesized that complaints concerning owners from poor socioeconomic regions would relate more to dog fighting, because of a known potential association with low socioeconomic status [32]. Finally, some research suggests that low socioeconomic status people are less likely to travel with their dogs, and therefore it is possible that low socioeconomic status is less likely to relate to dogs being left in a hot vehicle [33]. This is the third report in a series relating to the analysis of RSPCA Qld canine welfare complaint data [see also 27,28].

#### **2. Materials and Methods**

From July 2008 to June 2018, RSPCA Qld received 129,036 canine welfare complaints. Some involving more than one dog were recorded as multiple complaints sharing the same case number, while others were recorded as one complaint with multiple animals. To avoid sample bias due to multiple entries, we only retained the first complaint of case numbers with multiple entries, discarding 21,439 entries as a result. There remained 107,597 canine welfare complaints for this retrospective study. The data analysis was originally undertaken on the entire dataset and then repeated with the reduced number. Finding the complaint distribution and demographics to be similar, we opted for the reduced dataset to avoid pseudoreplication. Animal welfare complaints that fell within the geographical zone of responsibility of RSPCA Qld (determined by a Memorandum of Understanding between RSPCA Qld and Biosecurity Queensland, the Government Department tasked with the administration of ACPA) were investigated by RSPCA Qld inspectors. All other complaints were referred by RSPCA Qld to Biosecurity Queensland to be investigated by their inspectors. All complaints received by RSPCA Qld were included in this study, regardless of which authority investigated them.

All complaints were recorded in ShelterBuddy® (RSPCA, Queensland, Australia), the RSPCA Qld database. The following information was requested from the reporter of each incident at the time of taking the complaint: the number of dogs involved and their age, breed(s) (if known), the "complaint code(s)", suburb, postcode and date. All cases were investigated either by RSPCA Qld

inspectors (n = 100,432) or Biosecurity Qld inspectors (n = 7165). It is recognized that some of the calls, once investigated, were found not to relate to a breach of the ACPA or to a genuine welfare concern. However, the outcome data for complaints were not analysed in this research, which focused solely on the complaint calls received by RSPCA Qld.

Dogs were classified according to two broad age ranges, dog and puppy, based on reporters' interpretation. It was important to recognise that the information recorded from the complainant may be inaccurate or inaccurately interpreted, e.g., a small dog is commonly referred to as a puppy in Queensland. Records regarding breed and the number of dogs involved were based on either complainants' initial reports or comments from trained inspectors, again recognising inaccuracies with identification of the breed. The "complaint code" was selected by the staff member receiving the call or email from a drop-down menu of 18 possible complaints (Appendix A) [28]. Multiple "complaint codes" were able to be selected for each case according to the description of what was alleged to have happened to the dog(s), and each was treated as a separate code for analysis.

#### *2.1. Socioeconomic Scores*

Australia is spatially divided into regions by postcodes; our dataset included the postcode of the location of the dog being reported, which was taken as a proxy measure for the socioeconomic status of the owner. We also reviewed the Australian government's Socio-Economic Indexes for Areas (SEIFA) developed in 2011 [34], which rank postcode regions in Australia by the socioeconomic level of inhabitants. Four indices are assigned to each area to describe the local socioeconomic status: (1) the Index of Relative Socio-Economic Disadvantage, (2) the Index of Relative Socio-Economic Advantage and Disadvantage, (3) the Index of Economic Resources, and (4) the Index of Education and Occupation. Each index is ranked by decile, percentile, and score. Among these four indices, the Index of Relative Socio-Economic Advantage and Disadvantage (IRSAD) correlates well with the other three and is suitable for comparing the entire range of areas, and was therefore considered the most appropriate descriptor of the socioeconomic level of inhabitants of each postcode [34]. The IRSAD score is a weighted combination of selected indicators of advantage and disadvantage: household income, availability of internet connection, educational level, occupation, employment rate, property, mortgage, and health. Nationwide, the score is standardized with a mean of 1000 and a standard deviation of 100, with a mean, median and standard deviation in Queensland of 999, 1014 and 65 respectively. An area receives a score of 1000 if all of the above indicators are equal to the national average; the score for an area increases or decreases if the indicators are greater or less than the national average, respectively [34]. The index is positively associated with socio-economic advantage and negatively associated with socio-economic disadvantage, thus a region with a high IRSAD score is more likely to have people with high socioeconomic levels dwelling within it [34].

#### *2.2. Dog Breeds*

Any breed in our dataset that was recognized by the Australian National Kennel Council (ANKC) [35], New Zealand Kennel Club (NZKC) [36], American Kennel Club (AMKC) [37], or United Kennel Club (UKC) [38] was considered a recognized breed (RB) (see Appendix B list of recognized breeds). Any other reported breed in our data was considered an unrecognised breed (UB), including all crossbred dogs without any identified breed. If more than one dominant breed was listed, the first mentioned was used. For instance, Great Dane × Bull Arab was categorized as Great Dane (Appendix B).

To achieve a secondary representation of breed recognition, RB breeds were amalgamated into the following seven breed groups based on the breed inclusion categories of the ANKC: toys, terriers, gundogs, hounds, working dogs, utility, and non-sporting. Breeds not listed by the ANKC but recognised by the NZKC, AMKC, or UKC were categorized into one of the seven groups based on the description of each kennel club. Some breeds (e.g., Australian Koolie and Bull Arab) were listed by the council registration, it being an obligation of all dog owners in Queensland to register their

dogs with the local council. As a result, they were on the breed list (Appendix B), however they were not recognised as breeds by any major kennel club worldwide. Therefore, these breeds were categorized as unrecognized breeds, UB. If the breed description was left blank, the dogs' breed was considered unknown (n = 15,576/107,597), and these complaints were excluded from any data analysis related to breed factors. Our previous study suggested that, compared to specific breeds, breed groups and the dichotomization into RB and UB provided better agreements between the public and the trained RSPCA inspectors, and were therefore used for genotype identification in a public reporting system [27]. Therefore, this study used breed groups and RB/UB dichotomization rather than specific breeds for statistical analyses.

#### *2.3. Statistical Analysis*

Data was analysed using the statistical package Minitab® 17.3.1. (Minitab, LLC., State College, PA, USA). The distributions of IRSAD scores of both our entire study group and the RB/UB differentiation were not normal. Box–Cox and John transformations were used, but the assumption of normal distribution of residuals still could not be met. Therefore, a one-sample sign test was used to compare the median IRSAD scores between owners of dogs involved in an alleged incident in our dataset and the entire Queensland population as recorded by SEIFA. A Mann–Whitney test was conducted to compare the IRSAD scores of postcodes where there had been reports of RB and those of UB. As for the IRSAD scores of owners of dogs of the different breed groups, normality was assessed by visual inspection of residual plots. All complaints were independent but the assumption of equal variance could not be met based on the Bartlett's test (*p* = 0.018). Consequently, Welch's ANOVA followed by the Games-Howell pairwise comparisons were used to compare IRSAD scores between dogs of the different breed groups.

To determine factors influencing complaint codes, the IRSAD score, dog's age (dog or puppy), and breeds (UB or RB) were entered into eleven binary logistic regression models as fixed factors, using non-linear logit models with an alpha value for variables to enter the model of 0.15 [39]. Complaint codes were entered into the model as outcomes. Each case was independent from each other, and little multicollinearity was observed for all independent variables, with the variance inflation factors being less than 5. The linearity of the independent variable (the IRSAD score for the postcode of the dog) and the log-odds was assessed by plotting the independent variable and the log-odds fitted with a linear regression line. The assumption of linearity was considered met when the *P* value of a straight-line regression was less than 0.05. Eight complaint codes—Abandonment, Baiting/poisoning, Causing a captive animal to be injured/killed by dog, Dog fighting or other prohibited offence, Emergency relief, Keeping or using an animal for blooding/coursing a dog, Prohibition order breached, and Tail docking or other surgical procedure—did not fulfill the linearity assumption. Therefore, to fulfill the linearity assumption, Causing a captive animal to be injured/killed by dog, Dog fighting or other prohibited offence, Keeping or using an animal for blooding/coursing a dog, and Knowingly allowing an animal to kill/injure another were combined and categorized as a new code—Dog fighting. Baiting/poisoning, Cruelty, Prohibition order breached, and Tail docking or other surgical procedure were combined and categorized as a new code—Cruel act. Abandonment did not meet the linearity assumption statistically, but the graph was linear by observation, so the code was still used to construct a stepwise forward binary logistic model. Finally, the IRSAD score was removed from the regression model of Emergency relief because it did not meet the linearity assumption and the code was rarely cited in the past decade (0.01%, n = 8) [28]. Eleven stepwise forward binary logistic regression models were constructed to examine how different fixed factors (IRSAD score, dog's age and breeds) correlated with different outcomes (9 complaint codes and 2 combined complaint codes). Separate models were constructed for each code with the same input variable. In this paper, we focused on the relationship of IRSAD scores with complaint codes and breed factors. The relationship of other variables with the complaint codes have been reported separately [27,28].

#### **3. Results**

#### *3.1. Descriptive Statistics*

The median IRSAD score of owners being reported (median = 975) was significantly lower (*p* < 0.001) than that recorded for the population of Queensland (median = 1,014), and the Q1 and Q3 values were less (Table 1). Owners of reported UB dogs had significantly lower IRSAD scores (median = 970) than those reported owning RB dogs (median = 981) (*p* <0.001) (Table 2). Mean IRSAD scores for the postcodes of the different breed groups reported were significantly different (*p* <0.001) (Table 3), with gundogs (994 ± 63.0) being the highest, followed by toy (986 ± 64.3) and hound (984 ± 63.7), then non-sporting (980 ± 64.6) and working dogs (977 ± 63.0), terriers (977 ± 62.7) and utility dogs (976 ± 62.2).

**Table 1.** Descriptive analysis of Index of Relative Socio-Economic Advantage and Disadvantage (IRSAD) scores of the postcode of dogs in our study and those determined for the population of Queensland.


Q1: the first quartile; Q3: the third quartile.

**Table 2.** Index of Relative Socio-Economic Advantage and Disadvantage (IRSAD) scores of the postcode of owners of dogs reported to be recognised breeds (RB) and unrecognised breeds (UB).


Q1: the first quartile; Q3: the third quartile.

**Table 3.** Total numbers and Index of Relative Socio-Economic Advantage and Disadvantage (IRSAD) scores of the postcode of reported dogs in each breed group (*p* < 0.001).


\* Means not sharing a letter are significantly different (*p* < 0.05) by Games-Howell pairwise comparisons. SD: standard deviation.

#### *3.2. Complaint Codes*

The IRSAD scores of postcodes of dogs who were or were not reported for each complaint code significantly differed for the nine regression models (Table 4). The odds ratio (OR) was defined as a one unit increase in the IRSAD score leading to a corresponding x-fold decrease or increase in the odds of the cited event. The following codes were associated with dogs from postcodes with low IRSAD scores: cruel act (OR = 0.9994, *p* < 0.001), insufficient food and/or water (OR = 0.9981 *p* < 0.001), no exercise/confined/tethered (OR = 0.9979, *p* < 0.001), no shelter (OR = 0.9990, *p* <0.001), no treatment (OR = 0.9986, *p* < 0.001), overcrowding (OR = 0.9982, *p* = 0.002), poor dog condition (OR = 0.9974, *p* < 0.001) and poor living conditions (OR = 0.9996, *p* = 0.002). A single code was associated with dogs in postcodes with a high IRSAD score, hot animal in a car (OR = 1.0067, *p* < 0.001).



Not by binary logistic regression (alpha 0.15). assumption being logit binary logistic regression model. \* Dog fighting is a combined code including Causing captive animal to be injured/killed by dog, Dog fighting or other prohibited offence, Keeping or using an animal for blooding/coursing a dog, Knowingly allowing an animal to kill/injure another. + Cruel act is a combined code including Baiting/poisoning, cruelty, Prohibition order breached, Tail docking or other surgical procedure.

a

#### **4. Discussion**

#### *4.1. Socioeconomic Status of Dog Owners*

A better understanding of relationships between the socioeconomic status of dog owners and specific welfare issues could help to elucidate the reasons for the issues, as well as helping to target specific sectors of the population for education about dog welfare. Compared to the median IRSAD score across the entire Queensland state, our study group had significantly lower IRSAD scores, indicating that the reported dogs were in postcodes with inhabitants at a lower socioeconomic level. It is possible that dog owners generally come from postcodes with inhabitants at low socioeconomic status, but because not all welfare issues were more frequently reported in lower socioeconomic regions and one was more frequent in higher socioeconomic regions, we consider that there may be an important relationship between welfare issues and socioeconomic status.

The score differences contributed by the variables appear small, and the odds ratios for IRSAD scores in the regression model were close to one. This is probably because the distribution of IRSAD scores across different regions in Queensland is narrow, with an interquartile range of 71.1 and a standard deviation of just 64.8, compared with a range of 558. Also, our dataset only covers the coastal area of Queensland which is relatively homogenous in terms of the socioeconomic level of inhabitants. The difference between our median score and that in Queensland, 38.3, represents about 59% of one SD in that region. Since about 68% of values lie within 1 SD of the mean, it can be seen that the differences in IRSAD scores in the variables tested in this study are meaningful and reflect the range in values from a significant proportion of the total Queensland population, about 40% (68% × 59%). Such a small difference is important because it should enable us to predict differences in commonly reported dog welfare concerns across populations and regions in a large and relatively homogenous area. The majority of previous studies have focused on socioeconomics and animal abuse [20,22,29]. The relationship between socioeconomic level and other welfare concerns in relation to canines [6] has received little research attention. This study may bridge this gap by determining some of the factors that relate to specific welfare concerns.

A key feature of the IRSAD score is that it might positively relate to the financial circumstances of the owner, notwithstanding the previously mentioned concern that the owner may not be represented by the status of the entire postcode, since it includes the % of people whose annual household income is < AUS \$20,799 and whose rent is less than AUS \$166/week [34]. This could indicate some constraints on the part of the owner in providing for the welfare of his or her dog, such as provision of adequate food, shelter, or other resources. This is discussed later in relation to individual complaint codes. The IRSAD score is also positively associated with the inhabitants' educational level, a key component being whether the members of the household progressed past year 11 in school, assuming they were over 15 years of age. This could affect whether the owner has sufficient knowledge to care for his or her dog, providing suitable nutrition, for example. Important detractors in the IRSAD include unemployment and the percentage of employed people classified as "labourers" living in that postcode [34]. This could relate to whether the owner has sufficient financial resources to care for his or her dog. However, these results should be interpreted with caution because the IRSAD score relates to the entire postcode, which may include substantial variation within and between regions. Clearly the circumstance of the owner or owners may be different to that of other inhabitants in the postcode in question.

Broad correlations between socioeconomic status and human behaviour towards animals have been noted previously [6,40,41]. People with high socioeconomic levels are more likely to advocate for animal welfare and to volunteer for animals [40], whereas those with lower socioeconomic backgrounds are more likely to be involved in animal neglect and abuse [6]. Our findings are in agreement with these previous conclusions.

It is also important to remember that in this study we analyzed only reported data, i.e., alleged welfare issues, which may not all reflect actual cases. It is known that complainants will report a neighbour out of spite, misread a situation, or report actions which may not actually represent a breach of the ACPA. The negative correlation between the socioeconomic status of the dog's owner and the risk of most reported welfare issues may also be explained by the fact that individuals may be more prepared or likely to report welfare issues in low socioeconomic postcodes. Since household income, home ownership, and full-time employment were reported to negatively correlate with dog ownership, it is less likely that the higher number of reports in low socioeconomic regions is because people from low socioeconomic regions own more dogs [42]. However, the same study reported living in rural locations was associated with higher odds of owning a dog [42], and we found a tendency for low socioeconomic regions to overlap non-urban regions [43]. Therefore, the socioeconomic level and dwelling in rural regions may both increase the possibility of being reported; yet we could not validate these hypotheses as the IRSAD score is a generalization of postcode regions ignoring the within postcode variation. However, inconclusive results were reported in another study, that found that ones' personal and household income levels were not associated with the propensity of being reported for animal cruelty [44].

#### *4.2. Breed*

Not only can socioeconomic level of dog owners be related to animal welfare concerns, it can also be linked with breed factors. Reported cases involving UB dogs were potentially related to owners being socioeconomically disadvantaged compared with cases involving RB dogs. Similar results of breeds' predisposition to welfare concerns and socioeconomic levels were observed when we examined the RB dogs. Reported cases involving utility breeds, terriers, and working dogs were associated with the three least socioeconomically advantaged groups. Utility breeds, terriers, and working dogs were also more commonly reported for canine welfare concerns [27]. Although this study involved only reported but not confirmed cases, these results potentially support previous studies suggesting that people with lower income have higher risks of mistreating animals [20,22], which might be further associated with specific dog breeds, and linked to some breed-specific complaints [27]. Nevertheless, it is important to note that the differences in socioeconomic scores among different breed groups were statistically significant but small, and there may be some inaccuracies and biases when breeds were reported. Therefore, over-interpretation should be avoided.

#### *4.3. Complaint Types*

There were clear correlations between the socioeconomic level and complaint type being reported. For example, allegedly committing cruel acts was associated with lower socioeconomic level.

Previous research focusing on crime and animal cruelty has found negative relationships between the socioeconomic level and tendencies to commit crime, including being cruel to dogs and cats [22,45]. However, a study comparing animal cruelty between rural and urban regions found that rural residents mostly targeted cats rather than dogs [10], and another study investigating community demographics of animal cruelty reports found no differences among urban, town, and rural residents in the likelihood of being reported for animal cruelty [43].

Except for canine abuse, most complaints for which the owner was from a low socioeconomic background were neglect related. The aforementioned explanation is that people with lower socioeconomic backgrounds may lack the ability (e.g., money, space, or transportation) to manage animal care and welfare [6]. Moreover, this finding is in line with a previous study that people of lower socioeconomic status tended to moralize transgressions that did not cause obvious harm to animals [46,47]. Less affluent people may therefore be more likely to view an animal welfare compromise that is not overtly cruel as a moral but not a legal issue [6,46,47]. Consequently, there is a risk that people with lower socioeconomic backgrounds may tend to neglect the fundamental needs of their dogs, including failing to provide appropriate nutrients, adequate living conditions, and medical treatments, which increases the chance of them being reported. In this respect, the first of our hypotheses was supported, that complainants from relatively poor socioeconomic postcode regions would be more likely to complain about an absence of key resources for dogs. This may relate

to the factors included in the IRSAD score that are relevant to educational level. The IRSAD score of different postcodes across Australia is highly correlated with the Index of Education and Occupation score, with a Spearman's rank correlation of 0.85 [34]. This indicates a potential relationship between the low levels of education and neglect-related canine welfare concerns, either because people who are less educated do not consider deprivation as an act of neglect, or they are simply lacking in knowledge about the welfare and care of dogs [6,29].

Among these neglect-related complaints, the finding associated with insufficient (medical) treatments seems contradictory to our previous study [27]. This study reveals that insufficient veterinary treatments were less commonly reported in regions of lower socioeconomic status, which is supported by the fact that household income limits owners' access or willingness to provide veterinary care [48,49]. In addition, UB dogs were more commonly reported in lower socioeconomic regions. Consequently, it would be expected that UB dogs would be the subject of more complaints about poor veterinary care. However, according to our previous study, RB but not UB dogs were more likely to be reported with insufficient veterinary treatments [27]. The potential predisposition of RB dogs to a complaint about lack of veterinary care may be influenced not only by owners' socioeconomic status, but also by other factors. These factors may include morality [10], attitudes to the welfare of breeding dogs [50], human–animal bond [49], and registration rate [51], which can outweigh or confound the effects of socioeconomic level. For instance, people who are affluent but have less moral conviction may prefer to purchase an RB dog from a breeder rather than adopting an UB dog from a shelter, and the owners with less moral conviction may be less likely to bring their sick dogs to a veterinary clinic [49]. Besides, the potentially low registration rate of UB dogs [27] may encourage owners to abandon their dogs when medical care is required, leading to a reduced but inaccurate prevalence of UB dogs being reported for lack of treatment. However, these hypotheses cannot be confirmed in this study as they were all reported not confirmed cases.

Another inconclusive finding was that the difference in socioeconomic levels between reported cases citing and not citing abandonment was small. Dog ownership is positively correlated to household income [30,42,52]. However, it has been suggested that people with lower household income are less likely to relinquish their dogs, for financial reasons [14]. Therefore, it has been argued that other variables, including problematic behaviours [15,53,54], nature of the dog [15,22,53], human factors [53,55], and the human–dog bond [14,56] may also play important roles in determining the benefits of dog ownership. This finding again supports the previous assumption that other factors also critically influence some complaint types.

Similarly, we hypothesized that complaints about 'blood sports', such as 'Knowingly allowing an animal to kill / injure another' and 'Dog fighting or other prohibited offence' would be associated with lower socioeconomic backgrounds. However, the IRSAD score did not significantly differ between reported cases cited and not cited with this code. Previous research exploring financial aspects of dog-fighting in the UK has pointed out that this kind of 'blood sporting' was more popular among working-class men, as a way of life and an alternative expression of masculinity [32]. Nevertheless, a small proportion of middle-class people might also be involved as business owners or as a hobby [32], and thus increase the average socioeconomic scores. In addition, only a small number of cases were reported involving those alleged complaints, with even fewer being confirmed [28]; thereby, it might not be enough to test for statistically significant differences with validity.

Although most complaints were related to socioeconomic disadvantaged people, one complaint type was reported more commonly among more socioeconomically advantaged people. Those with relatively higher socioeconomic levels were more likely to be reported leaving their dogs unattended in a hot car. This finding partially supports the previous study that people with higher socioeconomic level, mainly living in urban areas, are more likely to own cars and take their dogs for outdoor activities [41], and thus have a greater chance of leaving their dogs alone in a car. In high socioeconomic regions and in high density urban regions it would be more likely that owners are reported when they left their dogs in a vehicle in a busy public area.

Although this study reports unconfirmed dog welfare complaints, the results reflect the major welfare concerns in dog populations in higher or lower socioeconomic background. Considering the relevance of different complaint reasons and different socioeconomic levels, intervention strategies for the prevention of animal neglect or cruelty should be directed differently in high and low socioeconomic regions. Studies of confirmed welfare issues are required. Interventions are recommended to be taken in lower socioeconomic areas to explore whether the high number of reports is driven by actual offenses or by higher public awareness. These results can also be used to increase public awareness and promote public education. For instance, councils of relatively higher socioeconomic regions are recommended to place more emphasis on enforcing that people do not leave dogs in a hot vehicle unattended. Councils of relatively lower socioeconomic regions can highlight information for owners on the basic needs of dogs (e.g., the amount of water and food consumption). Another important implication of this study is that it provides information regarding the correlation of socioeconomic backgrounds and the preference for dog types, which could help develop more tailored educational programs that target different populations.

#### *4.4. Limitations and Need for Future Research*

Several limitations were identified in this study. First, the dataset consisted of cases reported but not confirmed, so results only reflect a likelihood of correlations between socioeconomics and different types of welfare concerns in dogs. Besides, the socioeconomic data was acquired by linking to the postcodes where an alleged welfare concern occurred (used as a proxy for the owners' postcode), which appeared to be a generalization and might be an ecological fallacy. Therefore, the results should be cautiously interpreted. Future study could try to obtain a more direct measure of socioeconomic status, for example, the household income of each individual. Second, total numbers of residents in each post code are not accurately known. Therefore, we cannot calculate the exact prevalence of each welfare issue in different areas. If the prevalence in certain regions is particularly high, then sampling bias may occur. Third, breed recognition was based on comments made by complainants or trained inspectors, which may not be accurate. Finally, the data was obtained from populations within Queensland, and thus wider geographical generalization should be made cautiously.

#### **5. Conclusions**

This dataset was analyzed based on reported but not confirmed cases of canine welfare concerns, so the results reflect the tendency rather than fact. Results correlate the socioeconomic level with different dog breeds. The relationships between socioeconomic levels and different complaint types are also identified. Reported dogs of unrecognizable breeds came from postcodes with lower socioeconomic status to those reporting dogs of recognizable breeds. Among RB dogs, reports concerning utility breeds, terriers, and working dogs were more common than dogs reported in socioeconomically disadvantaged areas, but it is not clear to what extent these breed groups are more prevalent in these areas. People living in lower socioeconomic regions were more likely to be reported to be involved in canine welfare concerns, especially neglect-related complaints, and abusing dogs. In contrast, people living in higher socioeconomic areas were more alleged to leave their dogs unattended in a hot vehicle. This study provides detailed information which may help in the development of tailored strategies for different populations to combat welfare concerns in dogs. However, the differences of socioeconomic level were relatively small so the results should be interpreted cautiously. Finally, more risk factors and their roles in different complaint types should also be identified in order to give a better picture of canine welfare concerns.

**Author Contributions:** Conceptualization, C.J.C.P., H.Y.S. and M.B.A.P.; methodology, H.Y.S., C.J.C.P. and M.B.A.P.; software, H.Y.S. and C.J.C.P.; validation, H.Y.S., C.J.C.P. and M.B.A.P.; formal analysis, H.Y.S.; investigation, H.Y.S.; resources, M.B.A.P. and C.J.C.P.; data curation, C.J.C.P., H.Y.S. and M.B.A.P.; writing—original draft preparation, H.Y.S.; writing—review and editing, C.J.C.P., M.B.A.P. and H.Y.S.; visualisation, H.Y.S.; supervision, C.J.C.P. and M.B.A.P.; project administration, C.J.C.P. and M.B.A.P.

**Funding:** This research received no external funding.

**Acknowledgments:** We thank RSPCA, Qld for providing the database of canine welfare complaints, and RPSCA inspectors for the consultation.

**Conflicts of Interest:** Mandy Paterson is employed as the principal scientist by RSPCA, Qld, but none of the authors received any interest or financial support from people or organizations that inappropriately influenced this study.

#### **Appendix A**


**Table A1.** Description of each complaint code, alleging a welfare issue.

<sup>a</sup> Prohibition order-A prohibition order is given by the court when a person convicted of an animal welfare offense must not possess any or specific animal for a prescribed period of time [8].

#### **Appendix B**



**Table A2.** *Cont*.


**Table A2.** *Cont*.


**Table A2.** *Cont*.


**Table A2.** *Cont*.


**Breed Reported Associated Listed Breed Breed Group** Shar Pei cross Shar pei Non sporting Shetland sheepdog Shetland sheepdog Working dogs Shiba Inu Shiba Inu Utility Shih tzu Shih tzu Non sporting Shih tzu × maltese Shih tzu Non sporting Siberian husky Siberian husky Utility Skye terrier Skye terrier Terrier Sloughi Sloughi Hounds Small terrier cross Terrier Terrier Smithfield cattle dog Cross breed UB Soft coated wheaten terrier Soft coated wheaten terrier Terrier Spaniel Spaniel Gundogs Spanish water dog Spanish water dog Gundogs Spitz Spitz Non sporting

Spoodle Cocker spaniel Gundogs Staffordshire bull terrier American Staffordshire bull terrier Terrier Staffordshire bull terrier × labrador American Staffordshire bull terrier Terrier Staghound Staghound UB Swedish vallhund Swedish vallhund Working dogs Tenterfield terrier Tenterfield terrier Terrier Terrier Terrier Terrier Thai ridgeback Thai ridgeback Hounds Tibetan mastiff Tibetan mastiff Utility Tibetan spaniel Tibetan spaniel Toys Tibetan terrier Tibetan terrier Non sporting Timber shepherd Cross breed UB Weimaraner Weimaraner Gundogs Welsh springer spaniel Springer spaniel Gundogs Welsh terrier Welsh terrier Terrier West highland white terrier West highland white terrier Terrier Whippet Whippet Hounds White Swiss shepherd dog White Swiss shepherd dog Working dogs Wirehaired fox terrier Fox terrier Terrier Xoloitzcuintle Xoloitzcuintle Non sporting Yorkshire terrier Yorkshire terrier Toys

**Table A2.** *Cont*.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
