Weight loss outcomes are generally worse for dogs and cats with class II obesity, defined as > 40% overweight

In pet dogs and cats, adiposity is most-often estimated clinically using a 9-category body condition score (BCS), with BCS 9 equating to ~ 40% overweight. Animals that are more overweight (> 40%) are seen in clinical practice but are not appropriately depicted by descriptions in the existing categories. To determine whether being > 40% overweight has clinical relevance, this study aimed to compare the outcomes of weight management in animals that were > 40% overweight with those < 40% overweight. Records of dogs and cats attending a specialist obesity care clinic, where adiposity is determined using dual-energy X-ray absorptiometry (DXA), were reviewed. Animals were assigned to two classes (class I ≤ 40% overweight: 118/398 [40%] dogs and 68/116 [59%] cats; class II, > 40% overweight: 180/398 [60%] dogs and 48/116 [41%] cats) based on DXA results, and weight loss outcomes were compared. Fewer class II dogs obesity completed weight management than class I dogs (P < 0.001), rate of weight loss was also slower (P = 0.012) and lean tissue loss greater (P < 0.001). Compared with class I, cats with class II obesity lost more weight (P = 0.048) albeit over a longer period (P = 0.043) leading to greater lean tissue loss (P = 0.004). Approximately half the pets presenting to a specialist clinic were have class II obesity (> 40% overweight), and some weight loss outcomes are worse for these animals.


Baseline data classified by degree of obesity
Of the 298 dogs, 118 (40%) and 180 (60%) were classified as having class I and class II obesity, respectively (Table 1).Not surprisingly given the classification criteria, body fat percentage (Mann-Whitney test, P < 0.001; rank biserial − 0.69 [very large effect]) and percentage overweight (Mann-Whitney test, P < 0.001; rank biserial − 1.00 [very large effect]) were both greater in dogs with class II obesity, but there were no differences between classes for age, sex, neuter status, therapeutic diet used and the presence of co-morbidities (P = 0.754 to 0.884, effect size 0.00 to 0.05 [very small or small effects]).Of the 116 cats enrolled, 68 (59%) and 48 (41%) were classified as having class I and class II obesity, respectively (Table 2).Again, body fat percentage (Mann-Whitney test, P < 0.001; rank biserial -0.93 [very large effect]) and percentage overweight (Mann-Whitney test, P < 0.001; rank biserial − 1.00 [very large effect]) were greater in cats with class II obesity.Further, a greater proportion of cats with class II obesity were female (class I 24/68, 35%; class II 28/48, 58%; Chi-squared test, P = 0.037; Cramer's V 0.21 [small effect]), but there were otherwise no differences in age, the presence of co-morbidities or diet between classes (Chi-squared tests, P = 0.435 to 0.979, effect size − 0.14 to 1.00 [small effects]).

Outcomes in animals with class I and class II obesity
Outcomes of controlled weight reduction for dogs in class I and II obesity are shown in Table 3.Compared with class I obesity dogs (78/118, 66%), fewer dogs with class II obesity (80/180, 44%) completed their weight reduction programme (Chi-squared test, P < 0.001; Cramer's V 0.20 [small effect]).Further, dogs with class II obesity took longer to complete (Mann-Whitney test, P < 0.001, rank biserial -0.42 [very large effect]), likely because they lost a greater percentage of weight overall (Mann-Whitney test, P < 0.001; rank biserial -0.30 [medium effect]) at a slower average rate (Mann-Whitney test, P = 0.012; rank biserial 0.18 [small effect]).Dogs with class II obesity also attended more visits than dogs with class I obesity (Mann-Whitney test, P < 0.001; rank biserial -0.25 [medium effect]), but there were no differences in energy intake during weight loss, the number of times the diet was adjusted or the number of times weight loss stalled (Mann-Whitney tests, P = 0.217 to 0.460; rank biserial -0.05 to -0.09, very small effects; Table 3).
Outcomes of controlled weight reduction for cats in class I and II obesity are shown in Table 4. Like dogs, cats with class II obesity took longer to complete (Mann-Whitney test, P = 0.043; rank biserial -0.29 [medium effect]); however, whilst cats with class II obesity lost a greater percentage of weight overall (Mann-Whitney test, P = 0.048; rank biserial -0.27 [medium effect]), there was no difference in rate of weight loss between cats with class I and class II obesity (Mann-Whitney test, P = 0.126; rank biserial 0.21[medium effect]).Further, there www.nature.com/scientificreports/were no differences in the number reaching target, the number of visits attended, the number of times the diet was adjusted and the number of weight loss stalls between cats in classes I and II obesity (Mann-Whitney tests, P = 0.118 to 0.586; rank biserial -0.06 to 0.21, very small to medium effects; Table 4).

Changes in body composition in animals with class I and II obesity
Body composition data from both before and after weight reduction were available from 73/78 (94%) and 77/80 (96%) dogs with class I and II obesity, respectively, that completed their weight reduction protocol (Table 3).The magnitude of change in fat mass was greater in class II compared with class I dogs (Mann-Whitney test, P < 0.001; rank biserial 0.41 [very large effect]).The magnitude of lean mass change was also greater in class II compared with class I dogs (Mann-Whitney test, P < 0.001; rank biserial 0.61 [very large effect]), whilst a greater proportion of class I obesity dogs gained lean tissue (Chi-squared test, P < 0.001; Cramer's V 0.31 [medium effect]) compared with class II obesity dogs.It was possible that the dogs with class II obesity had lost more lean tissue simply because their percentage weight loss was greater.To examine this, a multiple linear regression model was constructed to determine the association between lean tissue change and obesity class, whilst accounting for the percentage of weight lost (Table 5; dog model).In this model, both class (regression coefficient -3.349, P = 0.007) and percentage weight loss (regression coefficient -0.394, P = 0.007) were independently associated with change in lean tissue mass.Body composition data from both before and after weight reduction were available from 39/44 (89%) and 19/22 (86%) cats with class I and II obesity, respectively, that completed their weight reduction protocol (Table 4).The magnitude of change in lean mass was greater (Mann-Whitney test, P < 0.001; rank biserial 0.57 [very large effect]) in cats with class II compared with class I obesity.However, there were no differences between classes in either change in fat mass (Mann-Whitney test, P = 0.135; rank biserial 0.29 [medium effect]) or the proportion of cats that gained lean tissue (Chi-squared test, P = 0.253; Cramer's V 0.17 [small effect]).As with dogs a multiple linear regression model was constructed to determine the association between lean tissue change and obesity class, whilst accounting for the percentage of weight lost (Table 5; Cat model).In this model, an association was identified between lean tissue change and percentage weight loss (regression coefficient -0.349, P = 0.002) but not with obesity class (regression coefficient -1.542, P = 0.409).

Discussion
The primary aim of the current study was to assign dogs and cats with overweight and obesity into two classes (I; ≤ 40% overweight; II > 40% overweight) based on body composition analysis by DXA, and then to compare differences in weight loss outcomes between classes.A further aim was to determine whether the proportion of dogs and cats with different obesity classes has changed over time, based on referrals to a specialist obesity care clinic.Overall, approximately half of all dogs and cats seen were classified as having class II obesity and, during controlled weight reduction, they lost weight more slowly and lost more lean tissue mass, than those with class I obesity.
In human medicine, obesity is subdivided into three classes (I, II and III) depending on degree of adiposity as defined by BMI 17 .As mentioned above, morbidity and mortality risk differ amongst classes 22 , with individuals having class III obesity being at greatest risk [23][24][25][26][27] .The use of classes, rather than historical descriptions (such as 'severe' or 'morbid'), also avoids the use of stigmatising language, not least given the prevalence and negative consequences of weight stigma both in society 28 and amongst medical professionals 29 .In the current study, the cut-point between the two classes (40% overweight) was chosen because this value signifies the upper limit of the 9-point BCS in dogs and cats 19,20 .Given this difference in assigning the cut-point, the two canine and feline Table 1.Baseline variables in dogs with class I and class II obesity. 1 Class I and II obesity defined as < 40% and > 40% overweight, based on analysis of body composition. 2Benjamini-Hochberg-adjusted P-values 68 reported for categorical data are from either Chi squared or Fisher's exact test; those for continuous data are from Mann Whitney tests. 3Effect size reported was Cohen's V for tests involving categorical data and rank biserial for continuous data.The magnitude of the effect size is reported according to the rules of Cohen 69 for Cramer's V and Funder and Ozer 70 for the rank biserial. 495% CI 95% confidence interval. 5Body fat percentage determined by dual-energy X-ray absorptiometry. 6Ideal weight determined from body composition analysis from DXA, as described in the "Methods" Section. 7HPMF high protein medium fibre, HPHF high protein high fibre.Significant values are in bold.To examine possible associations with morbidity, we compared differences in comorbidities between obesity classes.In contrast to humans, the proportion of individuals that had at least one comorbidity did not differ between animals with class I and class II obesity.However, we should be cautious in interpreting these results.Firstly, because the study was undertaken at a specialist obesity care clinic, the cases studied might not have been representative of pet dogs and cats attending primary care practices.This might explain why most animals studied had at least one comorbidity, with the effect that the statistical comparison was probably not meaningful.Further, many different comorbidities were present which were diverse in causes and consequences (Supplementary Data S1); grouping these comorbidities for the purpose of statistical analysis might have meant that genuine associations between obesity class and single comorbidities were missed.This was partially addressed in dogs by assessing orthopaedic, cardiorespiratory and dermatological disease separately; however, different diseases would still have been grouped into these single body system categories.Therefore, associations between obesity class and the presence of comorbidities would be better studied in epidemiological studies using larger, and more representative populations.A further limitation was the fact that we did not assess associations between obesity Table 2. Baseline variables in cats with class I and class II obesity. 1 Class I and II obesity defined as < 40% and > 40% overweight, based on analysis of body composition. 2Benjamini-Hochberg-adjusted P-values 68 reported for categorical data are from either Chi squared or Fisher's exact test; those for continuous data are from Mann Whitney tests. 3Effect size reported was Cohen's V for tests involving categorical data and rank biserial for continuous data.The magnitude of the effect size is reported according to the rules of Cohen 69 for Cramer's V and Funder and Ozer 70 for the rank biserial. 495% CI 95% confidence interval. 5Body fat percentage determined by dual-energy X-ray absorptiometry. 6Ideal weight determined from body composition analysis from DXA, as described in the "Methods" Section. 7HPMF high protein medium fibre, HPHF high protein high fibre.Significant values are in bold.class and other adverse health impacts known to be associated with overweight status, such as lifespan, quality of life and comorbidities [2][3][4][5][6][7][8][9][10] .Again, therefore, further studies would be required to assess the full impact on of obesity class on health and well-being.
In humans, outcomes of conventional weight loss interventions are similar amongst classes, with individuals typically losing ~ 5-6% of their starting weight, whatever their obesity class 30 .In contrast, some weight loss outcomes did differ between obesity classes in dogs and cats of the current study: although both cats and dogs with class II obesity lost more weight than those in class I, weight reduction protocols took longer overall.Further, the rate of weight loss was slower, and more visits were required in dogs, but not cats, with class II obesity.Longer protocols requiring more visits are likely to be more challenging for owners, given the time commitment required, whilst a slower rate of weight loss could also be discouraging, increasing the chances of non-compliance or early discontinuation.This might explain why, compared with class I obesity, a lesser proportion of dogs with obesity class II reached their target weight.The slower rates of weight loss and poorer compliance are likely to be the result of the known physiological adaptations that occur during weight reduction, which antagonise the progress and can also promote weight regain 31 .In humans, several mechanisms have been identified including alterations in circulating concentrations of appetite-related hormones (e.g.increased ghrelin and gastric inhibitory polypeptide; decreased leptin, peptide YY, cholecystokinin, amylin, insulin and glucagon-like peptide-1) 31 , compensatory changes to energy homoeostasis (decreased energy expenditure due to reduced body mass and enhanced metabolic efficiency) 32 , altered nutrient metabolism that can alter energy homeostasis 33 ) and subjective changes in appetite (e.g.increased perception of hunger 31 ).Although research is more limited, some of these factors have also been demonstrated to occur during weight loss in companion animals.For example, weight loss in dogs leads to an increase in ghrelin concentration and a decrease in leptin concentration 34 .Further, energy expenditure decreases during weight reduction in both cats 35 and dogs 36 with obesity, and maintenance energy requirements remain low even during subsequent weight maintenance 37,38 .
Whatever the underlying mechanisms, given that outcomes are worse in dogs and cats with class II obesity, future research should aim to develop treatments and strategies better tailored to such individuals.For example, therapeutic diets could be better formulated to mitigate lean tissue loss, perhaps, by altering protein and amino acid profiles or adding functional ingredients; in this respect, l-carnitine has effects on protein turnover and energy expenditure 39 , and is already included in therapeutic weight loss diets given its positive effects in promoting muscle mass 40 .Pharmaceutical agents could also be considered; licensed drugs were previously available for dogs, such as dirlotapide, which produced significant weight reduction in clinical trials 41 but had side effects, and was eventually withdrawn from the market.Alternatively, off-label use of the newer human obesity drugs could be considered, such as semaglutide, which produces significant weight loss and also improves health outcomes such as reducing cardiometabolic risk and increasing physical function 42 .If such drugs were similarly effective in cats and dogs, there would be more therapeutic options, not least in cases where progress on a therapeutic weight loss diet is slow.Until either novel diets or drugs become available, a pragmatic approach could instead be considered for dogs and cats with class II obesity, such as using partial weight reduction protocols 43 .With such protocols, the target weight set is deliberately greater than the ideal weight, with the intention of maximising any benefits, such as functional improvements and quality of life whilst, concurrently, reducing the chances of failure because of non-compliance and discontinuation 44 .Partial weight reduction protocols are more likely to succeed than complete protocols 43 , but are a compromise and some negative health consequences might remain; for example, given that animals remain in overweight condition even after the end of their protocol, there is still likely some increased risk of developing comorbidities [8][9][10] .
Despite the longer duration and greater number of visits, there were no significant differences between classes in the number of times the weight loss stalled or the number of times that a change in the diet plan was required.The number of weight loss stalls is a crude metric for the challenges faced by an owner during a weight loss plan because more weight loss stalls would be expected in more challenging plans.Similarly, poor compliance is likely when owners are finding the weight reduction process challenging and, in such cases, more diet changes might be expected.Therefore, the current results might suggest that, despite a longer and slower process, the day-to-day challenges were not different for owners.Arguably, however, these variables do not capture the full extent of challenges experienced by owners whose pet is undergoing controlled weight reduction.To explore this more completely, additional metrics would be needed including the amount of food-seeking behaviour and diary records of non-compliance with the diet.Further, since this was an observational study, causality cannot be assumed and, in fact, there might be inverse causality.In this respect, owners who are finding the process more challenging might be more resistant to changing the plan, for example, by reducing the daily food portion; this might then lead to a slower rate of weight loss, with the effect being a longer plan overall.
Compared with dogs with class I obesity, those with class II obesity lost a greater amount of fat during their controlled weight reduction plan which is, perhaps, not surprising given that their starting fat mass was greater.However, change in fat mass during weight reduction did not differ between cats in the two obesity classes.This finding might be explained by the variability in how much body fat mass changed and the fact that there was marked overlap between classes (class I -86 to -30%; class II -83% to -16%).Therefore, although a medium effect size was observed, the group sizes might have been too small to enable the detection of a statistically-significant difference between group medians in hypothesis testing.To address this, further work would be required where body composition changes during weight reduction are assessed in a larger population of cats with obesity.
Loss of lean tissue during controlled weight reduction has also previously been reported in humans 45,46 , cats and dogs 13,14,44 .A novel finding of the current study was that losses were greater in dogs with class II compared with class I obesity.Since previous studies have shown that lean tissue loss correlates with overall percentage weight loss 13,14,44 , this class difference might simply be due to those with class II obesity animals having lost more weight.Indeed, in cats, the class difference disappeared when percentage weight loss was accounted for using multiple regression.However, in dogs, the class effect on lean tissue loss remained even after adjusting for percentage weight loss.Possible mechanisms for the additional lean tissue loss seen in dogs with class II obesity include a difference in severity of insulin resistance along with differences in adipokine profiles.For example, in experimental murine studies of obesity-related insulin resistance, an increased loss of muscle mass occurs via insulin receptor substrate-1/phosphatidylinositol-3-kinase Akt pathway down-regulation 47 ; since adiponectin inhibits the muscle degradation that occurs via this pathway 48 , the decreased adiponectin concentrations that arise in individuals with obesity might be contributory.These results are interesting given that, in cats, preweight-loss adiponectin concentrations are negatively correlated with the amount of lean tissue lost during weight reduction 49 , although this has not been seen in dogs.Of course, differences in other adipokines might also be responsible, such as resistin, which has been implicated in impairment of myogenesis and maintenance of muscle mass in humans 50 .It was beyond the scope of the current study to explore differences in insulin resistance and adipokine patterns between the obesity classes, but this could be considered in a future study.
Whether the additional lean tissue lost in animals with class II obesity is the result of the overall percentage of weight lost (both dogs and cats) or other factors (dogs only), it is an important consideration; for example, loss of muscle mass is an independent predictor of mortality in human chronic diseases 48 , and whole-body protein catabolism is increased both in humans with type 2 diabetes mellitus and those with obesity 47 .As mentioned above, lean tissue mass decreases during weight loss using diet-based strategies in humans 45 , and the consequences of this have been described, including decreased metabolic rate and increased risk of injury 46,51,52 .A decrease in metabolic rate can make any weight loss hard to sustain, and this might be a reason for subsequent regain of weight 46 .Given the adverse effects seen in people, it is likely that excessive lean tissue loss during weight reduction in dogs and cats might have similar negative effects.For this reason, the authors recommend taking steps to limit lean tissue loss during controlled weight reduction, such as using partial weight reduction protocols.In a recent study in cats with obesity, such an approach limited the amount of lean tissue lost 43 .Table 3. Outcomes of weight loss in dogs with class I and class II obesity.Categorical data are expressed as number (%), whilst continuous data are expressed as median (range). 1 Class I and II obesity defined as < 40% and > 40% overweight, based on analysis of body composition. 2Benjamini-Hochberg-adjusted P-values 68 reported for categorical data are from either Chi squared or Fisher's exact test; those for continuous data are from Mann Whitney tests. 3Effect size reported was Cohen's V for tests involving categorical data and rank biserial for for continuous data.The magnitude of the effect size is reported according to the rules of Cohen 69 for Cramer's V and Funder and Ozer 70 for the rank biserial. 495% CI 95% confidence interval. 5Percentage weight loss and rate of weight loss expressed as a percentage of starting body weight. 6Average energy intake expressed as kJ or kcal per kg 0.75 of ideal body weight in kg per day. 7Change in fat mass, change in lean mass and percentage of weight lost as fat were determined by comparing differences in body composition before and after weight reduction, as described in the "Methods" section.Note that changes in lean mass could be positive or negative, with positive values suggesting that lean tissue increased during the weight reduction period.Significant values are in bold.Approximately half of all the animals in this study (60% dogs, 41% cats) had class II obesity, with the proportion remaining relatively stable over time in dogs but not cats, where the prevalence between 2014 and 2017 was greater than the prevalence between 2009 and 2013.Given that these cases were seen at a specialist obesity care clinic, it is unclear as to whether a similar prevalence occurs in the wider pet population.Nonetheless, the findings are important because their visual and physical characteristics are not appropriately depicted by the commonly used 9-point BCS system 19,20 .The current study used DXA to classify such cases, a technique that is not widely available, thereby limiting the generalisability of the current results to primary care practice.Even with investment in such equipment, the additional cost implications for owners of pets with obesity might limit its use.Therefore, for the concept of obesity class to be translated to primary care veterinary practice, simple clinical Table 4. Outcomes of weight loss in cats with class I and class II obesity.Categorical data are expressed as number (%), whilst continuous data are expressed as median (range). 1 Class I and II obesity defined as < 40% and > 40% overweight, based on analysis of body composition. 2Benjamini-Hochberg-adjusted P-values 68 reported for categorical data are from either Chi squared or Fisher's exact test; those for continuous data are from Mann Whitney tests. 3Effect size reported was Cohen's V for tests involving categorical data and rank biserial for continuous data.The magnitude of the effect size is reported according to the rules of Cohen 69 for Cramer's V and Funder and Ozer 70 for the rank biserial. 495% CI 95% confidence interval. 5Percentage weight loss and rate of weight loss expressed as a percentage of starting body weight. 6Average energy intake, expressed as kJ or kcal per kg 0.67 of ideal body weight in kg per day. 7Change in fat mass, change in lean mass and percentage of weight lost as fat were determined by comparing differences in body composition before and after weight reduction, as described in the "Methods" section.Note that changes in lean mass could be positive or negative, with positive values suggesting that lean tissue increased during the weight reduction period.Significant values are in bold.www.nature.com/scientificreports/assessments would be required.One option might be to use a different approach to assess adiposity, such as the body fat index (BFI) 17,18 .With this metric, visual and physical characteristics are subjectively assessed using a rubric, which enables a crude estimate of body fat mass to be made.In this respect, BFI scores of 30, 40, 50, 60 and 70 correlated with ~ 14%, ~ 30%, ~ 67%, ~ 100%, and ~ 167% overweight.Therefore, animals with a BFI of either 30 or 40 could be classified as obesity class I, whilst those with a BFI of 50, 60 or 70 could be classified as obesity class II.One downside of the BFI is that it is less well known by veterinary professionals compared with the 9-point BCS.Further, the visual and physical characteristics assessed differ in the systems used for dogs and cats, making the system more complicated than BCS.A second option for a clinical tool to estimate body fat mass is to use zoometry, again as previously described 17,18 .However, this technique is complicated, taking longer to complete and requiring greater patient co-operation, whilst marked variability can occur with tape measure measurements, adversely affecting accuracy 53 , not least when used in multi-veterinarian practices or in more challenging animals (aggressive or nervous).Such issues limit widespread acceptance of alternative techniques to the 9-point BCS, not least in busy primary care practices where the available time in consultations might be limited.Therefore, the third option would be to consider redesigning the existing 9-point BCS system, for example, by adding more categories for cats and dogs that are > 40% overweight.This would require further research whereby visual and physical characteristics of dogs and cats with class II obesity could be assessed to determine characteristics distinguishing these animals from those of existing classes (especially BCS 9).Any revised system would then need to be validated by comparing its performance with a gold-standard measure of body fat mass such as DXA.Of course, as with any new system, there might then be challenges with acceptance from veterinary professionals.A pragmatic approach could be to continue to use the existing BCS system, but also flag any individuals whose visual and palpable characteristics suggest that they are beyond the upper limit of the scale (i.e., by recording them as "9+ " or "above 9").One strength of the current 9-point BCS is that it can be used to estimate ideal weight 54 , because the relationship between body fat mass and BCS is approximately linear 19,20,55 .A limitation of such the pragmatic approach of adding a "9+ " category, is that estimates of ideal weight in that category would be problematic.For such cases, other methods would be required to estimate ideal weight, for example, by assessing historical weight records to identify a prior adult weight where the dog was in optimal body condition, which can then be used as an ideal weight.
As for any scientific research, there are limitations that warrant consideration, some of which have already been discussed, including the fact that the study population might not be representative of the general pet population.Owners who agree to be referred to such a clinic might well be more motivated and, as a result, outcomes might be more favourable than for animals in the general pet population.Further, there was variability among animals and the study population was relatively small (especially in cats), and this might have obscured the identification of relationships between the degree of obesity and outcomes of weight reduction.This was compounded by the fact that some data were missing, for example, post-weight-reduction body composition data in animals where DXA was not performed after weight reduction.This limited our ability to analyse changes in body composition during the weight reduction period and might have contributed to the failure to detect a significant difference in fat mass change between classes in cats despite a medium effect size.
In conclusion, the subclassification of canine and feline obesity into classes I and II has been described, with animals with class II obesity having worse weight outcomes than those with class I obesity.Based on such a classification, many pet dogs and cats presenting to a specialist obesity care clinic would have class II obesity and, therefore, not be well represented by the current 9-point BCS system.

Animals
All participating animals had been referred to a specialist obesity care clinic for dogs and cats (Royal Canin Weight Management Clinic, University of Liverpool, Neston, UK) for investigation and management of obesity or obesity-related disorders between October 2004 and December 2022.To be eligible, animals had to have had body composition measured by DXA and had to have reached an endpoint for their weight reduction protocol, in a similar manner to previous studies in cats and dogs 15,16 .Animals that reached target weight were classified as "completed, " whilst those that did not were classified as "stopped prematurely" (including euthanasia or stopping at the owner's request).Animals whose weight reduction programme was still ongoing in December 2022 were not eligible for inclusion.Finally, animals had to be above their ideal weight (based on body composition analysis, see below), could not have a comorbidity that would make weight loss contraindicated, and controlled weight reduction would reasonably be expected to improve their health.
The University of Liverpool Veterinary Research Ethics Committee (RETH000353 and VREC793) and the Royal Canin Ethical Review Committee (150720-55) both approved the study protocol.As part of the University approval, the nature of the procedures performed were considered and, specifically, whether they should be classified as experimental procedures.In this respect, all clinical procedures were conducted in accordance with relevant guidelines (e.g. standard operating procedures) and regulations.Further, the foods used were fed for the clinical benefit of the study animals, and were commercially-available therapeutic diets commonly used by veterinarians to manage obesity.As a result, in the ethical approvals granted (RETH000353 and VREC793), neither the clinical procedures used nor the clinical use of the therapeutic diets were deemed to involve animal experimentation and, therefore, fell outside the remit of national legislation (e.g. the revised Animals [Scientific Procedures] Act 1986).The study is reported in accordance with ARRIVE guidelines (https:// arriv eguid elines.org).

Measurement of body weight and body composition
Body weight was measured by electronic weigh scales, which were regularly calibrated using test weights (2-50 kg; guaranteed to be accurate to within ≤ 0.5%; Blake and Boughton Ltd., Thetford, UK).Body composition was analysed in all patients using the same fan-beam DXA (Lunar Prodigy Advance; GE Lunar; Madison, USA), calibrated on a weekly basis using a phantom supplied by the company, and which has previously been shown to have high precision for repeat analysis in dogs 12 .Patients were either sedated (if DXA alone was performed) or anaesthetised if required for additional procedures, and scanned in dorsal recumbency, as previously described 13,14 .Purpose-designed computer software (Encore 2004, 8.70.005;GE Lunar) was used for data analysis 13,14 .

Weight reduction protocol
Full details of the weight reduction protocol used have been published previously [13][14][15][16] .Briefly, at the first visit, patients were weighed, their BCS assessed and their body composition measured by DXA (see below).Health status was determined by routine haematology, serum biochemistry, free thyroxine measurement (in dogs) and urinalysis.If necessary, additional diagnostic investigations (e.g.diagnostic imaging, additional laboratory investigations) were performed to determine the status of any comorbidities.A tailored weight reduction protocol was then formulated for each animal, again as previously described 13,14 .Briefly, animals were fed dry or moist therapeutic diets appropriate for their species (Tables 6 and 7; Royal Canin, Aimargues, France), which were either high in both protein and fibre (HPHF) or were high in high protein with a moderate fibre content (HPMF).The diet choice depended on owner and animal preferences.
In dogs, the initial food allocation for the weight reduction protocol was determined based on ideal weight (as calculated by DXA), in a two-stage process; first, the metabolisable energy requirement for maintenance (a.k.a.maintenance energy requirement, MER) was estimated as 440 kJ (105 kcal) × body weight [kg] 0.75 /day; the degree of restriction for each dog was then individualised based upon sex (female < male), neuter status (neutered < intact) and other factors (i.e.presence of associated diseases that might affect energy expenditure such as orthopaedic disease).This initial food allocation was typically between 50 and 65% of MER at ideal weight Table 6.Average composition of the therapeutic diets used for weight reduction in 298 dogs with obesity.DM dry matter, ME metabolisable energy content, calculated using the National Research Council 2006 predictive equation based on total dietary fibre (TDF). 1 Obesity management, Royal Canin, 2 Satiety weight management, Royal Canin, 3 Obesity management, Royal Canin, 4 Satiety weight management, Royal Canin, 5 Expressed as grams per 100 g.In addition to advice about feeding the therapeutic diet, owners also received tailored advice on lifestyle alterations to assist the weight reduction process.This included a physical activity plan, tailored to owner circumstances, individual animal factors and the presence of comorbidities.Advice could include recommendations about play activity in both dogs and cats, but also walking, running, agility training and hydrotherapy, in dogs.
After the initial visit, animals were then reassessed every 7 to 21 days to have their body weight measurements taken, and changes were made to the dietary and exercise plan if necessary.In animals that reached their target weight, a final evaluation was conducted.Health status was determined based on physical examination, haematology, serum biochemical analysis and urinalysis.Body weight and body condition were recorded, and body composition was reassessed by DXA.

Determining ideal body weight and obesity class
For the purposes of this study, 'ideal weight' was defined as the weight at which body fat mass was determined to be optimal, assumed to be ~ 15-25% for cats 13,18 and 20-30% for dogs 12,14,17 .To estimate ideal weight, DXA measurements of lean mass, fat mass and bone mineral content (in grams), before weight reduction, were entered into a spreadsheet (Excel for Mac, version 16.71, Microsoft).This spreadsheet incorporated a bespoke mathematical formula for modelling body composition changes during weight loss 13,14 , which assumed a median proportional loss in tissue mass of 83% fat: 17% lean; the ratio used was based on analysis of all available body composition data from the specialist obesity care clinic using the same diets, same DXA machine and software.Once ideal weight had been determined, percentage overweight was then calculated using the following formula: Animals were then assigned to one of two obesity classes: class I (< 40% overweight) and class II (≥ 40% overweight).This cut-point was selected because it equates to BCS 9 in both cats and dogs 19,20,55,56 .

Determining changes in body composition during controlled weight reduction
Where body composition data were available from both before and after weight management (i.e. in most dogs and cats that reached their target weight), changes in body composition were assessed.Changes in fat and lean mass were determined from the following equation: Changes in fat mass were always negative, suggesting that adipose tissue was always lost during weight reduction; however, changes in lean mass could be positive or negative: a positive lean mass change was seen when the lean mass after weight reduction was greater than that before, and suggested some gain of lean tissue.

Determining sample size
Since the study was observational, and there have been no previous attempts to classify canine and feline obesity according to class, a sample size calculation was not possible.Instead, we aimed to include as many dogs and cats as possible from those that had attended the obesity care clinic during the timeframe.However, the numbers of both dogs and cats in the current study was greater than the numbers studied in previous publications from the same obesity care clinic 3,[13][14][15][16]43,56 .

Dataset, variables assessed and missing data
The dataset on which all statistical analyses were conducted is provided in the online supplementary information for the study (Supplementary Data S1).Continuous data are expressed as median and range, whilst categorical data are reported as a number (percentage).In addition to obesity class, baseline variables recorded were age, breed, sex, neuter status, comorbidities (classified according to body system affected), body fat percentage, the percentage overweight and diet used (classified according to the macronutrient profile).Further, to assess changes in the distribution of obesity classes over time, animals were assigned to one of four different year groups, covering the period of enrolment to the study: 2004-2008, 2009-2013, 2014-2017 and 2018-2022.Outcome variables recorded were number reaching target weight, duration, percentage weight loss, rate of weight loss, average energy intake during weight loss, number of visits, number of times the diet was adjusted (e.g. by changing the daily food portion or type of diet used), number of weight loss stalls (where there had either been no weight loss or weight gain since the previous visit), percentage change in fat mass, percentage change in lean mass and number gaining lean mass.
Post-weight-reduction body composition data from unavailable from 140 dogs and 50 cats that did not complete their weight reduction protocols, and a further 8 dogs and 8 cats that did complete, because follow-up DXA was not performed.Further, average energy intake data could not be calculated in 9 dogs and 8 cats that did not return after the first visit.Otherwise, there were no missing data for any other variable assessed (Supplementary Data S1).
Vol  67 .Except where indicated, two-sided P-values were adjusted to control the false-discovery rate using the Benjamini-Hochberg method 68 , with the level of significance being P < 0.05.

Statistical analyses
The distributions of all continuous data were first assessed by examining histograms, Q-Q plots and using the Shapiro-Wilk Test.Given that most datasets did not have a normal distribution, a decision was made to use non-parametric statistical methods.A Chi-squared test for trend was used to establish if there was any change in the prevalence of class I and class II obesity over time (e.g.across the four year groups), with the effect size determined by calculating Cramer's V alongside its 95% confidence interval (95% CI); these effect sizes were interpreted according to the rules of Cohen 69 with 3 degrees of freedom; therefore, values of 0.06, 0.17 and 0.29 denoted small, medium and large effects, respectively.The Kruskal-Wallis test was used to explore changes in percentage overweight over time further, by comparing animals in the four year groups.Post-hoc comparisons amongst groups were made using Dunn's test, with the false-discovery rate controlled using the Benjamini-Hochberg adjustment 68 .The effect size was estimated by calculating η 2 alongside its 95%-CI, whilst the magnitude of the effect was interpreted according to the rules of Cohen 69 , with values of 0.01, 0.06 and 0.14 denoting small, medium and large effect sizes, respectively.The Mann-Whitney test was used to compare continuous baseline and outcome variables between animals with class I and class II obesity, with the effect size determined by calculating the rank biserial and its 95% confidence interval (95% CI).Rank biserial effect sizes were interpreted according to the rules of Funder and Ozer 70 , with larger values (positive or negative) indicating larger differences between groups (tiny: < 0.05: very small: 0.05-0.10;small: 0.10-0.20;medium: 0.20-0.30;large: 0.30-0.40;very large: > 0.40).Finally, either the Chisquared test or Fisher's exact test (if one or more cells in the contingency table had an expected count of < 5) were used to assess categorical baseline and outcome variables between animals with class I and class II obesity.Once again, Cramer's V was used to indicate effect size and interpreted according to Cohen 69 , albeit with 1 degree of freedom; therefore, values of 0.10, 0.30 and 0.50 denoted small, medium and large effects, respectively.For diet, two comparisons were made: cats fed dry food exclusively with those fed either wet food exclusively, or a combination of wet and dry food; sex and neuter status were compared separately in dogs, but only sex was compared in cats (given that all were neutered).When assessing comorbidities in dogs, the presence of any comorbidity was assessed, as well as separate assessments for orthopaedic, cardiorespiratory or dermatological disease (the three most common body systems affected); given the smaller number of cats, comparisons were only made for the presence or absence of any comorbidity and presence or cardiorespiratory disease.No statistical comparisons amongst breeds were made because dogs were from many different breeds and the numbers within each breed were often small (Tables 1 and 2).
Finally, multiple linear regression was used to examine associations between change in lean mass and obesity class, whilst accounting for percentage weight loss.Model performance was assessed using adjusted R 2 and the associated P values, whilst several methods were used to ensure that model assumptions were met: first, normality of residuals was confirmed by visually-inspecting Q-Q plots and using the Shapiro-Wilk test; second, homogeneity of variance was confirmed using visual inspection of a plot of fitted values against the square root of the standardized residuals, and also with the Breusch-Pagan test; third, no influential datapoints were identified using Cook's distance; finally, absence of multicollinearity was confirmed by assessing variance inflation factors (VIF) and ensuring that all were < 4.

Figure 1 .
Figure 1.Percentage overweight in 298 dogs attending a specialist obesity care clinic stratified by year of enrolment (2004-2008, red; 2009-2013, green; 2014-2017, blue; 2018-2022, purple).The circles represent data from individual dogs or cats, thick horizontal black lines represent the median of each year group, whilst the upper and lower hinges of the boxes represent the inter-quartile range (IQR).The lower whisker represents the smallest observation greater than or equal to the lower hinge of the box minus 1.5 times the IQR; the upper whisker represents the largest observation less than the upper hinge of the box plus 1.5 times the IQR.The red dotted line depicts 40% overweight, meaning that animals below and above this line would be classified as classes I and II obesity, respectively.No time effect was evident for proportion of dogs in class I vs. class II obesity (Chi square test for trend, P = 0.358; Cramer's V 0.07, 95% CI 0.00-1.00[small effect]) and for the percentage overweight (Kruskal-Wallis test P = 0.490, η 2 − 0.002, 95% CI − 0.01 to 0.04 [small effect]).

Figure 2 .
Figure 2. Percentage overweight in 116 cats (b) attending a specialist obesity care clinic stratified by year of enrolment (2004-2008, red; 2009-2013, green; 2014-2017, blue; 2018-2022, purple).The circles represent data from individual dogs or cats, thick horizontal black lines represent the median of each year group, whilst the upper and lower hinges of the boxes represent the inter-quartile range (IQR).The lower whisker represents the smallest observation greater than or equal to the lower hinge of the box minus 1.5 times the IQR; the upper whisker represents the largest observation less than the upper hinge of the box plus 1.5 times the IQR.The red dotted line depicts 40% overweight, meaning that animals below and above this line would be classified as classes I and II obesity, respectively.Significant time effects were evident both for the proportion in class I and II obesity (Chi square test for trend, P = 0.025, Cramer's V 0.18, 95% CI 0.00-1.00[small effect]) and percentage overweight (Kruskal-Wallis test P = 0.020, η 2 0.09, 95% CI 0.00 to 0.20 [medium effect]), with post-hoc testing indicating that the main group difference was between cats in year groups 2009-2013 and 2014-2017 (P = 0.047). https://doi.org/10.1038/s41598-023-50197-ywww.nature.com/scientificreports/

Table 5 .
Multiple linear regression assessing the association between change in lean tissue mass and both obesity class and percentage weight in dogs with obesity undergoing controlled weight reduction.a Estimate and 95% confidence interval (95% CI) of the regression coefficient for the predictor variable; for obesity class, the coefficient represents the expected difference in percentage lean mass change for dogs with class II compared with class I obesity dogs; for percentage weight loss, the estimate represents the change in lean mass for each percentage of weight lost.b Model performance assessed by calculating R 2 adjusted for the number of predictors in the model.Significant values are in bold.

Table 7 .
5verage composition of the therapeutic diets used for weight reduction in 116 cats with obesity.DM dry matter, ME metabolisable energy content, calculated using the National Research Council 2006 predictive equation based on total dietary fibre (TDF).1Obesitymanagement,RoyalCanin, 2 Satiety weight management, Royal Canin,3Obesity management, Royal Canin,4Satiety weight management, Royal Canin,5Expressed as grams per 100 g.(as determined by DXA, see below).In cats, the initial energy allocation for the weight reduction protocol was 126-146 kJ(30-35 kcal)per kg of ideal weight (again, as determined by DXA, see below), with no adjustment for sex, neuter status or the presence of comorbidities.