Questionnaire design and implementation.

The questionnaire (S1 File) had five sections. Section 1 focused on detailed cat demographics. Section 2 featured questions related to the cat’s body weight and body condition. Three BCS measures were asked in this section: 1) participant-perceived BCS designated by participants based on descriptions in the questionnaire (‘BCS Owner’), 2) BCS that was determined by participants choosing from five different depictions of cat shape the one most similar to their cat’s shape (‘BCS Figure’), and 3) BCS determined by the cat’s veterinarians in the past year (‘BCS Vet’). The BCS of 1-to-5 for BCS Owner and BCS Vet were labelled as: ‘very underweight (BCS of 1)’, ‘somewhat underweight (BCS of 2)’, ‘ideal (BCS of 3)’, ‘chubby/overweight (BCS of 4)’, ‘fat/obese (BCS of 5)’. Section 3 contained questions associated with ownership, particularly with how the cat was fed, the lifestyle of the cat, and the interactions between the participant and the cat. Section 4 explored the participants’ attitudes towards feline O&O. Ten questions asked about participant attitudes, five about attitudes to feline overweight and five about attitudes to feline obesity (Table 3; question 30 in S1 File). In the questionnaire, the terms ‘chubby’ and ‘fat’ were used as proxies for overweight and obesity, respectively. Each question provided five options (strongly disagree, disagree, neutral, agree and strongly agree). Lastly, Section 5 sought general demographic information about the participant. Participants with more than one cat were asked to answer the questions regarding the first cat when their names were listed alphabetically.

The questionnaire used an online interface provided by SurveyMonkey (SurveyMonkey Inc., San Mateo, California, USA). The survey was open for two months from 1^st Nov 2016 to 31^st Dec 2016 and was promoted through various ways such as online posting, sending leaflets to veterinary clinics, animal charities and advertising the survey on social media. The advertisements did not mention overweight or obesity but did mention the aim of the study (S2 File). Ten AUD50 gift cards were advertised as the incentive to increase the response rate.

Data management.

Data cleaning and management were undertaken in Microsoft Excel (Microsoft Corp. Redmond, Washington, United States) and R version 3.3.0 (R Core Team) with RStudio interface (RStudio Team), facilitated by the ‘car’ [30] and ‘plyr’ [31] packages. All the analyses were conducted in RStudio.

A response was included in the analysis only if the participant answered at least one of the questions about the evaluation of the BCS of their cat. For Australian residential participants who provided their postcode, participants were classified as living in ‘urban’ or ‘rural’ areas by consulting information from a marketing website [32].

Two main analyses were conducted: the first examined the associations between the owners’ attitude towards feline O&O and the owner-reported BCS of their cats; the second investigated the risk factors for feline O&O and underweight by using multinomial logistic regression. The significance level was set at P<0.05 throughout this study unless indicated otherwise.

BCS (outcome variable).

The three candidates for the cat BCS outcome variable were BCS Owner, BCS Figure, and BCS Vet. Although BCS Vet was most likely to be close to the true BCS of cats, it could not be considered because many participants (n = 571, 41.1%) did not provide this information. To determine whether BCS Owner or BCS Figure was more suitable, the levels of agreement between BCS Vet and both BCS Owner and BCS Figure were evaluated separately by calculating weighted kappa using the ‘psych’ package [33]. Furthermore, the weighted kappa between categorised body weight and BCS Owner and BCS Figure were calculated for several breeds of cats whose ideal weight ranges were documented on a website named ‘Cat Owner Club’ [34]. BCS Owner [with three categories: underweight (BCS of 1 or 2), ideal weight (BCS of 3) and O&O (BCS of 4 or 5)] was chosen as the outcome variable for the analyses reported here because the values of weighted kappa between BCS Owner and both BCS Vet and categorised body weight from Cat Owner Club were higher than those between BCS Figure and these two variables.

The association of the participants’ attitude towards owner-estimated feline O&O with their cat’s body condition.

Answers for each of the ten questions were used to classify each of the participants as having an (a) approving attitude (towards overweight or obesity), (b) neutral attitude or (c) disapproving attitude. Using owner-assessed cats’ body condition (i.e., ideal and O&O) as the binomial outcome and the responses to ten questions about participants’ attitudes as explanatory variables, binomial logistic regression was conducted in R to examine their associations. Underweight cats were excluded because the attitude questions were not about underweight in cats. Multivariable logistic regression was not conducted because the explanatory variables were clustered in two groups and were not independent. Two questions: “Being chubby is a disease” and “Being fat is a disease” were excluded in the analysis due to the undefined state of disease for O&O in (veterinary) medicine [35].

Analyses of the risk factors for feline overweight and obesity and underweight.

Explanatory variable management. The explanatory variables included were grouped into ‘cat demographic-related factors (Table 1)’, ‘feeding-related factors (S2 Table)’, ‘activity-related factors (S3 Table)’ and ‘participant-demographic-related factors (S4 Table)’. Some potential confounders of participants’ demographics (i.e., age, gender, education level, being a veterinarian and animal-related profession) were included in the analyses (S3 Table). All the variables in Table 1 and S2 Table–S4 Table were included in the analyses except for neuter status, due to all the intact cats being ideal-weight. Explanatory variables were re-categorised if, in the contingency table against outcome variable, more than 20% of the cells had less than five cats and if any of the cells of contingency tables contained a zero. Re-categorisation was performed to ensure no compromise of the biologically meaningful inference, as described in footnotes for Table 1 and S2 Table–S4 Table.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Contingency tables of potential cat demographic risk factors for feline underweight or overweight and obesity with different body condition scores (BCS; 1 to 5) evaluated by 1,390 cat owners, based on data collected by an Australian-based online survey in 2016.

https://doi.org/10.1371/journal.pone.0234190.t001

Multinomial logistic regression analysis. Multinomial logistic regression was conducted with the ‘nnet’ [36] and ‘car’ packages [30] in R. Univariable analyses identified unconditional relationships of the outcome variable with explanatory variables, which were included for multivariable model selection if they had a P-value less than 0.20. Variables with more than 15% of data missing were not considered for inclusion in the multivariable analysis. Collinearity between the explanatory variables was tested by Spearman’s rank correlation coefficient with the ‘Hmisc’ package [37]. If a correlation coefficient between a pair of variables was greater than 0.7, only the variable with a stronger association with the outcome variable (i.e., smaller P-value in the univariable model) was retained for further analysis. A forward variable selection process was applied using the significant level as the criterion of inclusion. Pairwise interactions that might be biologically meaningful were evaluated in the model and, if significant, retained. Confounders were included in the final model if at least half of the coefficients of the variables in the final model, changed by more than 20% after the addition of the confounder in the model. Two separate models with binomial outcomes, underweight versus ideal weight and overweight versus ideal weight, with all the final explanatory variables included were fitted to conduct the Hosmer-Lemeshow test to examine the quality of the final multinomial model fit [38]. This test was accomplished with the ‘ResourceSelection’ package [39].

Reliability of the responses.

The reliability of the responses was estimated by Cohen's kappa with ‘psych’ package [33] by comparing the responses of two hunting-related questions. The question asked whether the cat hunted, and the reported frequency of hunting was re-categorised as ‘hunting’, ‘not hunting’ and ‘I am not sure’. The other question asked what prey the cat hunted. If the participants named any type of prey in the response, their cats were considered to be cats that hunted; otherwise, the cats remained in the same ‘not hunting’ or ‘I am not sure’ category. Responses that contained missing answers to any of the two questions were excluded.

Response.

Of the 1,469 questionnaires completed, 1,390 questionnaires included participants’ evaluation of the BCS of their cats and were thus integrated into the analysis. More than half of the participants heard about the survey from social media (691, 53.0%) and 29.6% (n = 386) of the participants obtained the survey from RSPCA Australia-related sources.

Demographics of the participants.

Of the 1,390 participants, 1,186 (89.0%) were female and approximately half were aged between 25 and 44 years (720, 51.8%). Eighty-nine (6.7%) participants did not live in Australia at the time of completing the questionnaire. For the other 1,242 (93.3%), the three states with the most participants were New South Wales (493, 41.0%), Victoria (264, 22.0%) and Queensland (230, 19.2%). Among Australian residential participants, 824 (68.6%) lived in an urban area and 377 (23.1%) lived in a rural area. More than half of the participants had a bachelor’s degree or higher (719, 54.0%) and 199 (15.3%) worked in an animal-related profession. Around half (684, 49.2%) of the participants owned only one cat, 450 (32.4%) owned two cats and 255 (18.4%) owned three or more cats. More participant demographics can be found in S4 Table.

Demographics of the cats.

The detailed demographics of the 1390 cats are displayed in Table 1. Nearly half of the cats were ≥3 to <11 years old (48.3%) and the majority were mixed-breed (80.0%). There were slightly more female cats (51.8%) than males (47.9%) and nearly all of them were neutered (97.8%). The most common source of the cats was a shelter, pound or charity (669, 48.1%), followed by a breeder (198, 14.2%), a friend or family member (184, 13.2%), street cat (138, 9.9%) and a pet shop (80, 5.8%). Summary statistics of variables related to feeding pattern and activity are listed in S2 Table and S3 Table, respectively.

BCS results.

The numbers of cats evaluated with ‘BCS Owner’, ‘BCS Figure’ and ‘BCS Vet’ are presented in Fig 1. More than two-thirds (954, 68.6%) of cats were considered to have a BCS Owner of 3, and BCS Figure tended to be judged lower than BCS Owner and BCS Vet. Of the 1,390 cats, the BCS of 571 (41.1%) cats had not been evaluated by a veterinarian in the previous year. The comparisons of the three BCS evaluated are shown in Table 2. Only 126 (9.1%) participants weighed their cats regularly, 318 (22.9%) weighed their cats from time to time, 593 (42.8%) were informed of the cat’s weight when visiting a veterinarian, and 349 (25.2%) did not monitor at all.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Body condition scores (BCS; 1 to 5) for 1,390 cats evaluated by different approaches, based on data collected by an Australian-based online survey in 2016.

https://doi.org/10.1371/journal.pone.0234190.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. The tabulated body condition score (BCS) evaluations and weighted kappa between different owner-reported BCS based on data collected by an Australian-based online survey in 2016 (n = 1,390).

https://doi.org/10.1371/journal.pone.0234190.t002

BCS Owner denotes participant-perceived BCS, as categorised by participants based on descriptions in words in the questionnaire; BCS Figure was determined by participants by selecting one picture of a cat whose shape was the most similar to their cats out of five pictures; BCS Vet was the BCS determined by veterinarians in the previous year and reported by the participants. ‘Other’ of BCS Figure includes the cats that were not nearby the participants while answering the questionnaire, so their BCS could not be compared to the figure and so a category according to the figure could not be recorded. ‘Other’ of BCS Vet indicates a cat that had not been evaluated by a veterinarian in the previous year.

Weighted kappa suggested that BCS Owner had better agreement with both BCS Vet and BCS from Cat Owner Club than BCS Figure (Table 2). Nevertheless, 20.8% (n = 44) of the cats evaluated by veterinarians to have a BCS4 were considered to have a BCS of 3 by owners, indicating that owners underestimated their cats’ body condition. On the other hand, 22.2% (n = 14) of cats evaluated by veterinarians to have a BCS of 2 were considered to have a BCS of 3 by the owners, indicating that owners were more likely to report ideal body condition scores. However, the prevalence of O&O estimated using BCS Owner was only slightly lower (24.2%) than the prevalence of 25.9%, calculated using BCS Vet. Both the values of weighted kappa with BCS from Cat Owner Club were low, especially the one with BCS Figure, whose 95% confidence interval covered 0, implying a non-effective agreement.

Descriptive statistics of the attitude towards feline overweight and obesity among the participants.

The descriptive results of the attitude-related questions are presented in Table 3. In general, more participants chose ‘disagree’ or ‘neutral’ than other options in these 10 questions. About a quarter of the participants strongly disagreed with two statements: ‘I think that it's fine for cats to be fat’ and ‘Being fat says that the cat has a quality life.’ However, a quarter of the participants (362, 26.0%) thought chubby cats were cute, and more than one-sixth (252, 18.1%) agreed that fat cats were cute. The same pattern appeared in the statements ‘Being chubby doesn’t equal unhealthy’ and ‘Being fat doesn’t equal unhealthy’.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Descriptive results of the questions exploring participants’ attitudes towards overweight (i.e., ‘chubby’ in the statements) and obesity (i.e., ‘fat’ in the statements) in cats in an Australian-based online survey in 2016 (n = 1,390).

https://doi.org/10.1371/journal.pone.0234190.t003

Multinomial model results.

The univariable results of 27 explanatory variables (including three confounders) that had a P-value less than 0.2 out of 47 variables investigated (including five confounders) are presented in Table 5. Among the 47 explanatory variables, two (home-made cat food and leftovers of human food) had greater than 15% of data missing (i.e., 16.6% and 16.3%, respectively). No collinearity between explanatory variables was detected.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Univariable results of the multinomial logistic regression model to evaluate risk factors for owner-assessed underweight and overweight and obesity (O&O) in cats with a P-value less than 0.2, based on data from 1,390 cats collected by an Australian-based online questionnaire in 2016.

https://doi.org/10.1371/journal.pone.0234190.t005

The final model results are presented in Table 6. Eight variables were statistically significantly associated with the outcome variable, and two confounders, education level and animal-related profession, were included in the final model. No tested interaction terms were significant. In the final model, the factors associated with an increased odds of feline O&O included being middle-aged, being mixed-breed, being fed twice daily, being fed more than four times a day or ad libitum, dry food being the major part of the diet, participants who determined the food quantity without considering the amount that cats ate, begging for food by cats, staying less often outdoors outside the property of the owner and apartment and townhouse dwelling. The factors associated with an increased odds of underweight in cats included being aged 11 years or older, being fed four or more times a day, and townhouse dwelling. The final model showed a good fit with the P-values of the approximated Hosmer-Lemeshow test >0.05 (i.e., 0.888 and 0.876 for models with binomial outcomes, underweight versus ideal weight and overweight versus ideal weight, respectively).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Final multivariable multinomial model for risk factors for owner-assessed underweight and overweight and obesity (O&O) in cats based on the model of 1,390 cats whose information was collected by an Australian-based online questionnaire in 2016.

https://doi.org/10.1371/journal.pone.0234190.t006

Age.

Cat body condition is known to be associated with age. Concurring with previous studies, our results showed the highest odds of O&O in middle-aged cats [6, 11, 13, 18–20, 64] and underweight in elderly cats [15, 20]. As with dogs and humans, there is a tendency for O&O to develop with age [66] because of a reduced energy requirement in cats as they age. However, a cat’s ability to digest fat (and potentially protein) also decreases with age, resulting in reduced energy absorption [67]. This outcome is particularly profound in cats over 12 years old and can be one of the reasons for the high frequency of underweight in old cats [67]. Some diseases common among old cats, such as hyperthyroidism [68] and diabetes mellitus [69], result in weight loss. Moreover, inappetence may reflect reduced senses of taste and smell in old cats, or the pain caused by dental and periodontal diseases [66].

Breed.

In the current study, mixed-breed cats had significantly higher odds of O&O than purebred cats but not pedigree cats. Mixed-breed has been shown to be associated with feline O&O in several studies [14, 19, 20, 64], although, in these studies, purebred cats were not separated into pedigree and purebred. As pedigree animals are more likely to be raised and cared for not purely for companion purposes, it is possible these cats are treated by their owners differently from purebred cats, resulting in the different odds of O&O in the current study. Corbee [70] reported a 45.5% prevalence of O&O in pedigree show cats, which is higher than most studies investigating the prevalence of feline O&O apart from Russell, Sabin [15]. Although their previous studies have also found a weak association between feline O&O with breed status [11, 17, 71], they all acknowledged that their sample sizes for purebred cats may have lacked power for meaningful comparison.

Daily feeding frequency.

In the current study population, most cats (61.0%) were fed twice a day. Compared to this group, cats fed once a day had lower odds of being overweight or obese, and cats being fed at least four times a day or ad libitum had higher odds. Relatively frequent daily feeding was associated with higher odds of feline O&O, although being fed three times a day did not show significantly higher odds than the odds of O&O when fed once a day. While several studies have explored the association of feeding frequency with feline O&O [8, 10, 14, 15, 17, 22, 72, 73], only Courcier, O'Higgins [17] reported significant results, in that feeding two to three times a day, but not once a day, was associated with higher odds of O&O than feeding ad libitum. In studies that have specifically explored associations between feeding ad libitum and feline O&O, inconsistent results have been found. Cave, Allan [10] reported a non-significant association, whereas Russell, Sabin [15] found a higher odds of O&O in cats fed ad libitum of canned food but not dry food. It is possible that feeding frequency may be ineffective as a causal factor of feline O&O because, to estimate the total amount of food eaten by cats, other factors such as the amount of food per meal and cats’ feeding behaviours must also be considered. Our results also showed that feeding four or more times a day occurred more in underweight cats. The frequent feeding of these cats is likely to be an owner's response to the underweight status instead of the cause of underweight.

Food type and feeding quantity.

We examined many types of food as risk factors for O&O and underweight in cats but found that dry food was the only risk factor for O&O. Dry rations have been thought to increase the risk of O&O due to their increased energy density. Although most studies that examined the associations of food types with feline O&O have reported non-significant results, two prospective studies, albeit from the same study population, have reported dry food as a risk factor for feline O&O [9, 22]. Among owners who answer the question of whether treats/snacks were fed to the cats, 57.3% more than half of the cats in the current study population were given treats/snacks (S2 Table). Although feeding treats and/or table scraps have been reported to relate to feline O&O [6, 9, 15], this association did not emerge in the current study. It is worth noting here that, in contrast to the current study, two of the studies that reported this association did not account for other potential risk factors associated with O&O in the analysis [6, 9, 15].

The amount of food fed to the cats as guided by their apparent appetite was shown to be associated with a lower BCS in the current study. Similarly, in another study, cats with owners who followed the instruction from pet-food companies were reported to have higher odds of O&O than those whose owners determined the amount of food based on their cat’s appetite [11]. However, although this practice seems to prevent O&O, it might also increase the odds of underweight in cats, as shown by our results.

Begging for food.

To the authors’ knowledge, this is the first study to investigate the feline behaviour of begging for food as a potential risk factor for feline O&O. Cats that begged often or always had more than three and five times the odds of being overweight or obese, respectively, than cats that never begged. It appears that, although the owners govern the provision of extra food to their cats, the behaviour of begging is persuasive enough to be a risk factor. Furthermore, besides sometimes using food as a reward, cat owners may sometimes misinterpret attention-seeking behaviours as begging for food and consequently over-feed their cats. This owner response can thus reinforce the behaviour of begging for food among cats [74]. Interestingly, the frequency of the behaviour of begging for food fitted the model better than the frequency of owners giving-in to cats’ begging. Therefore, it is possible that the participants underestimated the frequency with which they surrendered to begging cats. Owners should be advised that most attempts to extinguish positively reinforced behaviour are met with a so-called extinction burst whereby the behaviour becomes more frequent before it disappears [75]. Knowing this proximate outcome helps to fortify owners’ resolve to ignore escalated begging. In the current study, begging frequency was related to only O&O but not underweight, indicating that begging is associated with only excessive energy intake but not with the cat’s current energy requirement.

Stay outdoors outside the property.

In the current study, cats ‘often spending time outside the owner’s property’ had more than three times lower odds of O&O than the cats ‘staying less often outside the owner’s property’. Although many previous studies have examined the associations between feline O&O and outdoor access [11, 17, 20, 22, 76] or time spent outdoors [9, 14, 15], only Rowe, Browne [22] and Teng, McGreevy [20] reported significant (negative) associations. The outdoor environment not only extends home ranges of cats with outdoor access [77, 78], it also offers them more stimuli and more opportunities to encounter prey. Both these interactions and roaming increase cat energy expenditure and decrease the risk of O&O. However, roaming outdoors also increases the chance of instances that may compromise cat health and welfare such as infectious disease, road traffic accidents and attacks by dogs [74]. Furthermore, predation of wildlife by cats is a severe conservation issue worldwide [79–81]. Possible solutions for cat owners to increase cat physical activity without being released outdoors include a generously sized backyard enclosure and indoor enrichment.

Dwelling type.

As confinement seems to be associated with O&O in cats, it is not surprising that cats living in an apartment or townhouse have higher odds of being overweight or obese. Scarlett and Donoghue [13] also reported apartment dwelling as a risk factor for feline O&O. However, cats living in a townhouse were shown to have higher odds of being both underweight and overweight or obese than cats living in a house. The results seem to contradict each other and warrant further investigation. Instead of reflecting any effect of dwelling type, these results might reflect attributes of the cat owners living in different dwelling types that were not measured in this study.