This is the largest study to date exploring risk factors for DM in dogs in primary-care practices. It is the first to look at data from across the whole of the UK, and to explore associations between initial management methods for DM and survival. Dogs > 8 years of age, female entire dogs, male neutered dogs, WHWT and Border Terrier breeds in particular, dogs documented as obese, or having a concurrent diagnosis of hyperadrenocorticism or pancreatitis were all associated with an increased odds of DM diagnosis. Variables associated with an increased hazard of death after diagnosis included dogs that were ≥ 10 years of age at diagnosis, entire, previously on glucocorticoids, having had a BG level > 40 mmol/L at diagnosis, or dogs that did not start insulin treatment.
The prevalence of DM in the current study (0.26%) was calculated for dogs aged 3 years or above, and therefore may not be directly comparable to the higher prevalence values of 0.32–0.36% reported by other UK and Australian studies assessing the overall population dogs within a primary-care setting [1,2,3]. However the current study may offer a more accurate representation of the wider UK dog population, as previous studies were limited to only insured dogs [1], or only data from England [2]. Similarly, prevalence estimates of 0.64–1.33% reported by studies of hospital-based populations [11, 36] are likely to be biased due to the referred source population [37], and may not accurately reflect the wider dog population. In the current study, only 5.9% (24/409) of incident cases were referred for advanced management, highlighting the importance of studies from primary-care practice given that referral centres may be missing almost 95% of DM cases from the general dog population.
The annual incidence risk in dogs aged ≥3 years in the current study was estimated at 0.09% (95% CI: 0.08–0.09%). The study by Fall et al. (2007) on a Swedish insured dog population reported the cumulative proportion of dogs developing DM by 12 years of age as 1.2%, approximating to an annual incidence risk of 0.1% [38]. It is very difficult to compare these two studies due to the different source populations and methodologies for calculating annual incidence risk. Human T1DM had a reported average annual increase of 2.8% worldwide between 1990 and 1999 [10], and previous studies have suggested that the incidence of canine DM may also be increasing [11, 12].
Diagnostic testing information was available for 95.8% (392/409) DM cases in the current study. Of these dogs 32.8% (134) were ketotic at diagnosis which, having been assessed via urinary dipsticks, is likely to be an under-estimate because this test will not detect all types of ketone bodies [39]. This relatively high prevalence of ketosis at the time of diagnosis in primary care practice has not previously been reported. The same percentage of dogs (32.8%, 134) either presented with or developed cataracts within 3 months of DM diagnosis. The current study did not discriminate between diabetic and non-diabetic cataracts. However, some cases that developed cataracts may not have been documented in the clinical notes, suggesting that this figure could be an underestimate. Previous studies have reported 50% of diabetic dogs developing cataracts within 6 months [40, 41]. Although diabetic cataracts can be managed surgically, this is not always affordable for owners and ocular health was cited as a contributory reason for euthanasia in 14.2% (33/233) of all euthanasia decisions.
Age was strongly associated with the odds of diagnosis of DM. Consistent with other studies [2, 11], dogs over 8 years were at increased risk, with those aged 10 to < 13 years having over 7 times the odds of DM diagnosis compared to dogs aged 3 - < 8 years. No association was detected with sex, which is consistent with other studies [2, 14]. However, there was a strong association with the combined sex-neuter variable. Similar to previous findings, neutered males had almost twice the odds [2, 3, 11], and entire females had three times the odds of developing DM compared to entire males. Where there is a large entire female population, such as in Sweden, the increase in female cases of DM is thought to be a reflection of dioestrus diabetes [16]. However, despite the evidence between hormonal changes in dioestrus and the development of diabetes [20, 26, 42], overall the current study found no significant difference between entire and neutered females in developing DM. Instead, it suggests that entire males may be “protected”. Cross-sectional studies in humans have reported that men with lower testosterone levels have an increased T2DM risk [43]. In spontaneous mouse models of T1DM, such as the non-obese diabetic mouse, females are predisposed to diabetes, an effect thought to be mediated through testosterone-driven effects on the microbiome [44]. Although the pathogenesis of DM in dogs is heterogeneous, one mechanism by which male entire dogs may be relatively protected is via increased testosterone. However, this hypothesis may only be relevant to a proportion of diabetes cases, and was not investigated directly in the present study.
Breed associations were consistent with previous findings [3, 13, 15, 16], and add weight to the evidence that DM has a genetic component [7, 13]. To identify new genes and potential treatment targets in canine diabetes, understanding which breeds are genetically protected from DM is just as important as identifying those with a predisposition. In an aim to explore the effect of breed on both predisposition and protection in the current study, breeds with ≥10 dogs within cases and controls combined, and/or breeds with ≥5 DM cases were included as individual breeds within the ‘breed’ variable. Other breeds falling outside this definition were combined as ‘purebred other’. This enabled these individual breeds to be evaluated within the multivariable logistic analysis, and after adjustment for other variables and confounders on breed, aiming to provide a more accurate understanding of associations compared to most previous studies that use univariable analysis only. However, despite 1205 dogs being taken forward to multivariable analysis, this categorisation was still under-powered to evaluate breeds adequately where there were very few or no cases, despite the breed itself being relatively common. In this respect, this study was unable to identify all breeds at high risk of developing DM, and the findings focus on the breeds that were commonly represented in the UK during 2016.
The current study identified that WHWTs and Border Terriers were “at risk”, having approximately 3 times the odds of DM compared to crossbreds. Samoyeds are frequently over-represented in DM cases in other studies and although there was only one Samoyed present in this case control study, it was a DM case. Breeds with reduced odds of DM compared to crossbreds included GSD and Shih-Tzu, as well as SBT, a breed consistent with suggestive findings in other studies [2], but not previously significantly associated with lower DM risk. Interestingly the current study contained 10 English Springer Spaniels and 10 Pugs, but there were no cases documented for either of these breeds. Pugs have not previously been associated with a decreased odds of DM, whereas English Springer Spaniels have been associated with both a reduced odds of DM in the UK [13], and an increased odds of DM in Australia [3]. This suggests that different genetic subpopulations of English Springer Spaniels with different susceptibilities to DM may exist across these two geographical regions. This may be a useful area for further research into the genetics of DM in dogs.
Dogs documented as obese were associated with 2.7 times the odds of DM diagnosis (95% CI 1.63–4.52, p < 0.001). Despite no clear biological reason for obese dogs to be prone to DM due to insulin deficiency [24], this finding adds to evidence from other studies that obesity may be a risk factor for the disorder [2, 26], potentially by causing insulin resistance. The present study relied on subjective, unprompted recording of obesity by the veterinary professional, therefore under-reporting was likely, particularly with regards to controls, because cases had more detailed histories/examinations. Diet has also been associated with DM in dogs [26], and because obesity is often associated with poor dietary control and limited exercise [45], another explanation is that “obesity” may be acting as a proxy for other associated risk factors. It is clear that more research is required to establish the exact link between obesity in dogs and DM. Interestingly obesity was not associated with, nor confounded by, prior glucocorticoid treatment. The latter was associated with roughly double the odds of DM, which is likely to reflect insulin resistance caused by these drugs [46].
Concurrent hyperadrenocorticism and pancreatitis were both strongly associated with DM diagnosis, consistent with several other studies [2, 3, 17, 27]. The results should be interpreted with some caution because the numbers of controls with these conditions in this study were low, albeit consistent with the prevalence of these diseases (approx. 0.02–0.04%) in the wider UK dog population [47]. The temporality, and causation, between pancreatitis and DM is also difficult to determine [32], and interestingly only 13% (6/46) of the pancreatitis diagnoses in the current study clearly preceded DM. In total pancreatitis was documented in an unprompted way in 11.3% of cases, similar to the 11.5% reported in a previous UK primary-care study [2], but less than the 17.7% reported in DM cases in primary-care practice in Australia [3], and the 19% reported by an Italian referral study [34]. It is likely that pancreatitis is under-reported in the current study for both cases and controls due to non-specific clinical signs, and lack of definitive diagnostics being performed. Further research is required to fully understand the interplay between these two diseases.
Where recorded, the most common management methods for DM in the first 3 months were home or practice BG curves (70.6%, 255/361), followed by spot BG measurements (68.4%, 247), even though the latter is considered unreliable for monitoring [48]. Only 4.2% [15] of dogs were managed by spot BG alone, suggesting this method is primarily used as an augmentation to other management techniques.. Identification of management techniques and diagnostics currently employed by primary-care clinicians by this study provides a benchmark against which individuals and clinics can compare their own processes and practices.
Median survival time for all dogs with DM was 15.6 months (95% CI: 10.4–20.0 months). This estimate is likely to be negatively skewed by inclusion of dogs where DM management was not realistically attempted. To account for this, an MST was calculated for all dogs surviving at least 7 days post diagnosis, estimated at 20.2 months (95% CI: 16.6–24.7 months). The all dogs MST estimated in this study is consistent with the 17.3 months reported by an earlier VetCompass study [2], but differs substantially from the reported MST of 2 months from a population of insured Swedish dogs [16], and of 32 months for dogs presenting to a referral hospital in Italy [34]. This may reflect differences between countries, or that dogs presenting to a referral hospital are more likely to have motivated owners and access to gold standard level of care. Successful treatment of DM requires substantial owner commitment, and given that 92.5% of the dogs in the current study died due to an owner’s decision to euthanise, it is clear that MST is strongly influenced by a variety of owner-related factors such as their finances, lifestyle or perception of the condition.
Age at diagnosis, neuter status and breed have all previously been associated with DM survival [2, 16], and were also associated in the current study. Dogs aged ≥10 years and Cocker Spaniels had twice the hazard of death compared to dogs 3 to < 8 years and crossbreds respectively. Conversely, Border Collies and neutered animals had a lower hazard compared to crossbreds and entire animals. The differences in survival between breeds may reflect genetic differences involved in the pathogenesis of the disease. Interestingly previously studies have found a female predisposition to DM in Border Collies [13, 16], suggesting dioestrus diabetes may be more prevalent in this breed. In the current study 92.3% (12/13) of the Border Collie cases were females, with 6 of these were entire and 6 were neutered at the time of diagnosis. The increased survival found in this breed could therefore be partially related to survival associated with dioestrus diabetes cases, where remission may be possible following ovariohysterectomy. Whilst the clinical signs associated with canine DM, such as polyuria and polydipsia are relatively consistent, the underlying pathogenesis is relatively heterogeneous. More detailed and consistent clinical phenotyping at the time of diagnosis e.g. measurement of pancreatic inflammatory markers, would enable improved disease classification, allowing further insights to be gained into the relationship between underlying pathogenesis and survival.
Overall, neutered animals had a lower hazard of death once diagnosed (hazard ratio (HR): 0.56, 95% CI 0.42–0.78, p < 0.001) which is consistent with the findings of a previous UK primary-care study [2]. There is no clear reason whether this is biological or due to neutering acting as a proxy for some other measure such as the owner’s ability and willingness to treat DM.
Insurance status and concurrent pancreatitis have previously been associated with survival in primary-care practice [2]. Pancreatitis was cited as a cause of death in 7.3% (17/233) of euthanised cases, but in the current study did not identify an association between survival with insurance status, or the diagnosis of concurrent pancreatitis or hyperadrenocorticism at the time of DM diagnosis. A recent referral study similarly did not identify association with pancreatitis and survival [34], suggesting that pancreatitis, where recognised in the 3 months before or after diagnosis, may not significantly affect survival.
Other variables associated with survival in the current study included BG level at diagnosis, insulin treatment, and glucocorticoid treatment. Dogs with a BG reading > 40 mmol/L at diagnosis had a hazard ratio of 2.75 (95% CI: 1.35–5.57, p = 0.004) compared to dogs with a BG reading < 20 mmol/L. Higher BG levels at diagnosis may reflect “sicker” dogs or those in diabetic ketoacidosis, yet neither hospitalisation, or being ketotic at diagnosis, (a proxy measure for diabetic ketoacidosis) were associated with BG levels or survival. A referral study of incident cases reported that BG levels in untreated dogs were not associated with survival in multivariable analysis [34]. Although this was a smaller study and limited to a referral caseload, these differing findings to the current study may reflect intrinsic differences between typical treatment options and survival in a referral versus primary-care setting. In primary-care practice, early identification of DM, before BG levels rise > 40 mmol/L, may be important for this population of dogs to have a better chance of stabilisation and survival. Dogs starting insulin had a significantly reduced hazard of death (HR: 0.08, 95% CI 0.05–0.12, p < 0.001), consistent with the fact that virtually all dogs are insulin dependent by the time of diagnosis [9]. There was evidence from the current study that previous glucocorticoid treatment increased the hazard of death (HR: 1.83, 95% CI 1.20–2.80, p = 0.005). This may result from increased difficulties in managing DM concurrently with other disorders that require glucocorticoid treatment.
The presence of ketotis at diagnosis was not adversely associated with survival, and diabetic ketoacidosis was mentioned as a reason for euthanasia in only 3.4% of cases. Despite diabetic ketoacidosis potentially being a life-threatening condition, it appears not to be a driving factor for survival. This may reflect the confidence of primary practitioners in attempting treatment of the condition, rather than opting for euthanasia, or an under-reporting of the condition. The latter is likely to be true in this study where urine strips to identify ketones were used as a proxy for diabetic ketoacidosis at diagnosis, and where further investigations to diagnose diabetic ketoacidosis at the point of euthanasia was often not undertaken or reported in the clinical notes.
The study included information on a variety of management methods that were used to manage/monitor DM during the first 3 months following diagnosis, including BG curves at the practice, recommended diet change, spot BG and fructosamine measurements. At a univariable level, several of these methods were associated with increased survival. However, because these management methods were highly correlated with each other, and with insulin treatment, it was inappropriate to retain them all in the final model. Insulin treatment was retained as this had the greatest impact on the HR. Because dogs that survived longer had a longer duration of exposure to the chance of receiving management methods, it was aimed to minimise this survival-bias effect by restricting observations on management to the first 3 months post diagnosis. As many of the management methods were associated with increased survival, it can be argued that contact with the practice within these first few months is more important for survival than the type of management technique per se. This contact is likely to be a proxy for owner commitment and compliance, as well as the availability of support from the practice, and it has been argued that assessment of these on long term survival is more important than considering specific patient characteristics [49]. Further studies into survival of DM cases should consider ways to capture this element of owner compliance.
There were a number of limitations to this study. As it relied on reviewing of retrospective Electronic Patient Records (EPRs), inconsistencies and inaccuracies within these may have led to either missing data, or misclassification for variables such as obesity/medications/concurrent conditions. Errors relating to these inconsistencies were more likely to occur with controls than cases because the latter group were more likely to have higher counts of veterinary visits and investigations for concurrent conditions. The number of dogs within each breed category was dependant on the popularity of that breed within the UK within 2016. This meant there was insufficient power to detect potentially significant differences within breeds where there were low numbers, for example the inability of this study to detect Samoyeds as high risk for developing DM. Additionally, although misclassification of DM was thought unlikely because this is a routine primary-care veterinary diagnosis, a veterinary diagnosis of pancreatitis or hyperadrenocorticism in DM cases may have been misclassified due to similarities in clinical signs or difficulties in interpreting diagnostic tests.