This is the largest study to date using primary-care veterinary data that reports on the common disorders in UK dogs. The study placed specific focus on exploring age, sex and neuter associations with disorder occurrence. At a precise level of diagnostic precision, the most commonly recorded disorders were periodontal disease, otitis externa, obesity, overgrown nail(s) and anal sac impaction. Whilst at a grouped level of diagnostic precision, the most common groups were dental disorder, skin disorder, enteropathy and musculoskeletal disorder. Awareness of the most common disorders of dogs can assist efforts to prioritise health reforms in dogs at a species level [45, 46]. Differing associations between categories of sex, neuter and age were reported for many common disorders, suggesting that these demographic features are important factors that need to be considered when exploring the epidemiology of these disorders and during the application of epidemiological information into health reforms. The overall findings can assist veterinarians and owners to prioritise preventive care and to facilitate earlier diagnosis of common disorders within dogs. The information on associations with age, sex and neuter provides additional contextual background to the complex world of disorder occurrence and can support targeted health controls for these subsets of dogs.

A smaller previous UK study using primary veterinary clinical data from 2009 to 2013 reported the most common disorders in dogs as otitis externa, periodontal disease, anal sac impaction, overgrown nail(s) and degenerative joint disease [7]. These findings are largely in line with the current study, though the current study reports a slightly higher prevalence of obesity and lower prevalence of osteoarthritis. Although this could indicate a genuine change in disorder prevalence over time, these differences are more likely to reflect methodological differences between the studies. Compared to the 3,998 dogs in the earlier report, the larger sample size of 22,333 dogs in the current study offers greater current precision. In addition, the methods used to extract disorder data from the clinical records have also advanced considerably during the intervening years [47], meaning that the current study was more highly powered and technologically enabled than the previous study. The ranking of the most common disorders in the current study that relied on veterinary clinical records differs substantially to results based on other data resources. Analyses of pet insurance records in Sweden have reported that skin and gastrointestinal disorders feature highly in dogs [48, 49]. A questionnaire-based survey of UK owners of dogs registered with the KC reported lipoma (4.3 %), skin cyst (3.1 %) and allergic skin disorder (2.7 %) as the most common disorders in dogs, possibly reflecting the prioritisation of the personal concerns of owners as well as true prevalence [40]. Based on a telephone survey, the most common disorders in dogs in the US were reported as musculoskeletal, dental, and gastrointestinal tract or hepatic disease [50]. Differences in the ranking of the most common disorders between these various data sources suggest that the information resource and data extraction methods can have a substantial impact on the results. Comparison between results from series of studies based on standardised core data collection and analysis methods is therefore more likely to offer safer inference between studies. This conclusion suggests the value of larger research programmes using cohort data collection nationally for more repeatable and comparable study results [13].

There are currently concerted efforts to identify key breed health issues by comparing disorder prevalence and risk between individual specific breeds and an appropriate comparator group such as crossbreds, other specific breeds [51] or all remaining dogs [52]. An application of the results from the current study would be to provide comparator disorder prevalence data on the wider dog population that could act as a baseline for the generation of breed-specific predispositions and protections. For comparative inference with the greatest reliability, breed specific data should be extracted from the same information resource using the same methods as the comparator group [52]. The findings from such comparisons can be used to develop breed specific health plans that prioritise common, severe or long duration disorders; an example of such breed health plans is shown by the UK Kennel Club’s ‘Breed health and conservation plan’ project [37]. Identification of differing predisposition and protection profiles for common disorders within and between breeds can promote improved understanding of the wider impacts from morphological and behavioural diversity selectively introduced into modern dog breeds. Awareness of these predispositions offers the prospect that co-ordinated health activities at a breed level may mitigate some of these negative welfare impacts [9, 53]. However, it is worth noting that direct comparison of prevalence values between breeds risks falling into the trap of univariable comparisons that do not account for confounding effects from other variables that may be associated with the disorder risk and that may differ between the breeds (e.g. age, sex, neuter status) [54]. The nature of such confounding is explored later in this discussion.

Strategies to improve canine health can be focused on dogs overall to reduce the welfare impact of common disorders such as obesity [55] or can be targeted to the specific needs of individual breeds [37]. The diversity of canine breeds, each with its own breed club structure that holds their own unique list of perceived priorities, makes the breed-focused approach understandable. And, indeed, this breed-focused approach is still a major plank of current health initiatives in dogs [37]. However, this breed-centric approach often tends to prioritise mitigation efforts on disorders that are either predisposed or perceived to have high genetic components within a breed above those that are common and modifiable, especially within the lifetime of the affected animals [12, 36, 56,57,58]. Especially for disorders that have high severity and duration, greater welfare gains may result from even modest reductions in the frequency of common disorders compared with even large proportional reductions in rare disorders [12, 36]. Although the current paper does not report on disorder risks within individual breeds, the current results could be useful as a baseline on general health issues in dogs to reframe future health reforms towards mitigation of more common disorders.

Veterinary expertise in managing and preventing common disorders offers many welfare benefits for dogs but it has always been challenging to prioritise allocation of teaching time across the breadth of topics needed within veterinary undergraduate curricula [59]. The findings of the current study enable visualisation of the majority component of a veterinarian’s daily workload and could assist in refining undergraduate veterinary teaching curricula to better equip new graduates with the necessary breadth of day-one skills [60,61,62]. Assuming that optimal allocation of undergraduate teaching times should take consideration of overall disorder prevalence, it is possible that the teaching time allocated for common disorders may currently be disproportionately short and that student education on the clinical management of these common disorders may rely excessively on the hidden curriculum or on experiences gained during extramural studies [63,64,65,66]. There is a risk that this scatter-gun approach may generate new graduates with widely varying beliefs about what constitures ‘best practice’, while also potentially allowing some important clinical topics to fall through the educational net with inadequate coverage. Similarly, the current findings can provide a picture of the typical clinical workload in primary-care practice for persons considering entering the veterinary profession or even for veterinary undergraduates contemplating the directionality of their future careers [67].

Access to reliable overviews of the main disorders managed in primary care practice can assist logistical exercises aimed at ensuring adequate resources for veterinary care at a national level. For example, the World Small Animal Veterinary Association has recently published a “List of Essential Medicines for Cats and Dogs” [68]. The list aims to support veterinarians in providing acceptable preventive care and treatment for the most frequent and important diseases in dogs and cats by facilitating medicines availability, drug quality, use and pharmacovigilance. Essential medicines were selected based on consideration of disease prevalence along with other factors including public/animal health relevance, clinical efficacy and safety, and comparative costs and cost-effectiveness. The value of such logistic exercises that require access to overall disease prevalence data has been especially evident at a national level during the recent UK plans to leave the European Union [69] and at an international level during the Covid19 pandemic [70]. Generation of ongoing and more detailed datasets describing the overall disease burdens in companion animal species looks set to become an increasingly valuable activity over the coming years as countries becomes ever more influenced by activities and events outside their own borders [71, 72].

Sample size estimation (power calculation) is a critical design component for any research project [73]. Failure to consider this step could lead to inclusion of too few animals with an under-powered study that misses significant differences that truly exist in the target population or, conversely, inclusion of too many animals can lead to wastage of resources and to ethical issues [74]. Paradoxically, a major challenge to sample size estimation for many studies in dogs has been access to reliable population-based prevalence data. This may partially explain the personal experience of the authors over many years that a large proportion of otherwise good epidemiological research has been published without an accompanying a priori sample size estimation and therefore offered limited inference, especially when negative findings are reported. As an additional support to improve future epidemiological study design for disorder studies in dogs, the current study provides generalisable results on disorder prevalence from a large cohort of dogs under UK primary care that can be used as the basis for sample size estimation (power calculation) [75].

Confounding describes the “mixing of effects” wherein effects from an exposure of interest (e.g. breed) on an outcome (e.g. a disorder) are conflated with the effects of another factor (e.g. age) that distorts the true relationship between the exposure and the outcome [76, 77]. Confounders may mask a true association so that it is missed or, conversely, generate an apparent association between the exposure and outcome even when no real association between them exists [78]. There are several approaches to dealing with the confounding problem, including randomisation, exclusion, matching and by using appropriate analysis [76]. Although randomisation reduces the necessity for prior consideration of potential confounding, the other three methods are reliant on a priori causal consideration of both measurable and unmeasurable variables as potential confounders [78, 79].

The current study placed special focus on exploring associations between age, sex and neuter status with common disorders in an effort to assist future research to better understand and interpret these variables as potential confounders. The current results suggest important confounding effects for age, sex and neuter status in many common disorders of dogs. Among the 70 most common precise-level disorders, male dogs had increased odds for traumatic injury, ear disorder, brain disorder, musculoskeletal disorder, behaviour disorder, foreign body, upper respiratory tract disorder, and adverse reaction to drug. Conversely, bitches had higher odds for urinary system disorder, incontinence and claw/nail disorder. Some of these sex associations have previously been identified whereas the current study presents novel findings for others. It is worth noting that these latter novel findings should be treated with caution until supported or refuted by future confirmatory analyses. Disorders with previously reported associations with sex include otitis externa, aggression and seizure disorder in males [80,81,82] and urinary tract infection, mammary mass, urinary incontinence and cruciate disease in females [42, 83,84,85,86].

To add further complexity to the effects of sex on disorder occurrence, the current study also identified associations between neuter status and several disorders. Neutered animals had increased odds for obesity, dental disorder, musculoskeletal disorder, incontinence and behaviour disorder. Conversely, entire animals had higher odds for female reproductive disorder, male reproductive system disorder, underweight and parasite infestation. Some of these associations have been previously reported, such as obesity, osteoarthritis and urinary incontinence in neutered dogs [87,88,89]. However, it is often stated that ‘association does not imply causation’ and the dangers of making this leap from association to assuming causality are especially valid when trying to interpret the neuter status associations reported in the current paper [90]. Neuter status is a time-varying variable, with all dogs starting life as entire. Neutering itself is generally irreversible so that the proportion of neutered dogs rises with age, assuming neutering does not shorten life. Associations between neuter status and disorder therefore become heavily confounded by age. For example, although neutered dogs had a much higher prevalence of periodontal disease compared to entire dogs (17.34 % versus 8.56 % respectively), the median age of dogs with periodontal disease was quite old at 7.54 years. Since these periodontal disorder cases were, on average, older, they had more time to be neutered, which might account in part for this association. In addition, the current study was a cross sectional analysis with neuter status recorded at the date of final available record and therefore it is unknown whether the disorder preceded or followed the neutering event for each individual dog. To fully explore the effects of neutering on disease risk, cohort study designs are needed whereby each dog is followed over time from puppyhood, taking note of the dates of neutering and disease occurrence [76]. However, to date, such cohort studies have been rare in veterinary studies on companion animals due to their complexity as well as temporal and financial costs, especially for disorders that may occur many years after the neutering event [89].

The current study reported the median age of affected dogs at the end of 2016 for each of the common disorders. These values allow assessment of the typical age profile for dogs affected with these disorders. It is noticeable that disorders diagnosed in dogs aged over 9 years largely include degenerative and neoplastic disorders, such as osteoarthritis, heart murmur, skin mass and lipoma, whilst disorders diagnosed in dogs under six years largely include disorders associated with infection, allergy, behaviour, trauma or dietary indiscretion such as otitis externa, pyoderma, vomiting/diarrhoea, conjunctivitis and claw injury. To date, there has been limited published evidence on the age profiles of dogs affected with common disorders. However, the current findings are in line with a Korean report in which young dogs (< 1 year) most commonly presented for preventive medicine, diarrhoea, vomiting and infectious diseases while older dogs (over 10 years) presented more commonly with heart disease, kidney disease, Cushing’s disease, and mammary tumours [39]. Similar to the confounding effects described above, the results of the current study suggest that age should routinely be considered as a potential confounder in risk factor analyses for common disorders. For example, the median age in the UK varies widely between common dog breeds: French Bulldog 1.3 years [28], Bulldog 2.3 years [25], Chihuahua 2.8 years [27], Pug 3.0 years [33], Miniature Schnauzer 3.8 years [32], Rottweiler 4.5 years [34], German Shepherd Dog 4.7 years [29] and West Highland White Terrier 7.8 years [35]. A direct comparison of disorder profiles or prevalence between these breeds that did not account for age could spuriously suggest predispositions to typical disorders of young dogs in breeds such as the French Bulldog and Bulldog and predispositions to disorders of older dogs in breeds such as German Shepherd Dog and West Highland White Terrier. The current study highlights the complexity that is introduced into disorder occurrence from their associations with age, sex and neuter status and emphasises that careful consideration should be given to confounding variables when planning, designing and interpreting disorder studies in dogs [91].

Application of veterinary clinical data for epidemiological analyses have some important limitations that have been previously reported [13, 27, 42, 43]. Additional limitations for the current study included a tendency of veterinarians to commonly record presenting signs in lieu of formal biomedical diagnostic terms e.g. heart murmur rather than degenerative mitral valve disease, which prevents allocation of these true underlying diagnoses to the relevant specific disorders. Whilst this might underestimate the prevalence of specific disorders, the current results do reflect current primary-care practice more accurately. The process of disorder diagnosis and management in primary care practice is complex, with multiple factors such as client finances and expectations affecting the perceived need or importance for gaining full or precise diagnoses. These factors will therefore also impact the results reported in the current paper which should be interpreted as the prevalence of disorders diagnosed as opposed to the true prevalence of the underlying disorders in this population of dogs. The disorder prevalence values reported in the current study reflect the diagnosis levels in the current population of dogs. However, differing confounding effects from age, sex, neuter and other profiles in alternative populations could result in different prevalence values in these other populations that share the same inherent disorder propensities. The current study shows results from an array of association tests with the aim of detecting patterns of overall association. The provision of p-values may assist the reader to interpret these statistical results but it should be noted that these p-values did not include adjustments to account for multiple testing such as the Bonferroni correction [92]. Consequently, this elevates the possibility of Type I error (false positive findings) such that individual test results should be interpreted with caution [93,94,95]. Associations between sex and neuter with the individual disorders from the current study should be treated as exploratory, and as useful hypothesis generators, rather than as confirmatory. The current study reports disorder prevalence (i.e. the proportion of dogs with the disorder of interest during 2016) rather than incidence (i.e. the proportion of dogs that developed the disorder of interest for the first time during 2016). Prevalence studies are biased towards reporting disorders with longer duration (e.g. generally chronic conditions such as osteoarthritis, obesity) compared with disorders of short duration (e.g. generally acute conditions such as conjunctivitis, otitis externa) [36]. The study extracted the first listed presenting sign for disorders that were not ascribed a formal biomedical diagnosis. This term was reported in the precise-level diagnosis results. Given that many underlying pathologies present with multiple common presenting signs (e.g. gastroenteritis may present with vomiting, diarrhoea or malaise), this approach may have segregated these underlying pathologies across several terms at the precise-diagnosis level. The current study aimed to overcome this limitation by merging these terms into higher levels of abstraction in the grouped-level diagnoses.