Dazzled by Science! Part Four: Databased Research

Dr. Lana Kaiser


March 22, 2022



Last time we talked about surveys in research, now we’re going to grapple with databased research. A database is an organized collection of electronic data that can be accessed and analyzed. There are hundreds, likely thousands, and maybe millions of  academic, medical, veterinary medical, government, insurance, and multispecies DNA databases that can be “mined” for research. With the development of electronic records in university veterinary teaching hospitals, specialty and primary practices, data on multiple species, breeds, diseases, and treatment can be relatively easily accessed and analyzed.


One thing to consider when looking at databased research is the study population. Is it a biased or skewed sample, or does it attempt to represent the population at large? For example, studies that retrospectively examine disease outcomes from a university veterinary hospital database likely represent a select group of animals, whose people choose (or were referred) and could afford to take their animal to a tertiary referral center. Thus, this population of dogs may not represent the “general dog population,” but rather a select group of a specific breed. Retrospective analysis always carries the risk that pertinent information is not included in the record. Some researchers try to avoid this issue by contacting the dogs’ owner, but owners’ recollections may be colored over time. University veterinary hospital database records are generally created from records written by veterinary students, interns and residents, and the records may or may not have undergone a thorough review by the attending veterinarian.




Researchers from the UK used a private veterinary practice database to test the hypothesis that when compared to heavier dogs, lighter dogs have an increased probability of receiving drug therapy for undesirable behaviors. They used the VetCompassTM Programme database that includes de-identified electronic patient records from primary-care veterinary practices in the UK. This database includes four veterinary hospital groups, most data came from one group. For this study undesirable behavior was defined as “any behaviour pattern with evidence in the clinical records that it was troublesome to at least one human associated with the dog including members of the veterinary team, trainers and groomers as well as the owners, friends or strangers who interacted with the dog.” That is an extremely broad definition of undesirable behavior. Dogs with organic disease were excluded. The 2013 database included 104,212 dogs under veterinary care in the UK. They calculated to detect a difference in  medication treatment of undesirable behaviors between small (<10 kg) and large  (>10 kg) dogs, they would need records from  4,847 dogs of smaller bodyweight and 19,385 dogs of larger bodyweight. It is unclear how they decided that > 10 kg was a “large dog,” and <10 kg was a small dog, or what group the 10.0 kg dog belonged to.


There were 404 dogs prescribed medication by the private practice veterinarian for 413 undesirable behaviors. The authors provide extensive description of the variables examined (age, sex, weight, breed, categories of undesirable behaviors, drugs prescribed etc.) and categorize the behaviors at both a specific and general level (nervous aggression vs aggressive). They used a variety of modeling techniques to identify associations between variables, and provide an exhaustive description of the relationships between variables. The odds ratio was used to determine the association between variables (OR, a statistical method to identify the strength of association between two variables: OR < 1 the variables are not associated; OR >1 the variables are associated.) They determined the estimated one year period prevalence of undesirable behaviors treated with medication was 0.4% (95% CI 0.35–0.43). When compared to crossbred dogs, three breeds were more likely to have a drug prescribed for undesirable behaviors: Toy Poodle (OR 2.75), Tibetan Terrier (OR 2.68) and Shih-Tzu (OR 1.95). Although these are small dogs, they failed to prove their hypothesis, size of dog did not predict prescription of medication for undesirable behavior. Increasing age was associated with increased odds of drug-prescribed undesirable behaviors, and both neutered males (OR 1.82) and intact males (OR 1.50) had increased odds compared with intact females. In the discussion they review all the associations and describe why or why not this data may fit with the current understanding of canine behavior, undesirable behaviors, management and treatment.


There is an overwhelming amount of data in this manuscript, but I am left with the sense that sometimes data mining is an end to itself. The hypothesis, unproven by the study, was based on two previous studies by the authors that using questionnaires to assess behavior (CBARQ and Dog Mentality Assessment) and a third study from Spain with a similar questionnaire, suggesting that small dogs may have more perceived behavior issues. The volume of data makes it difficult to ferret out data that is cogent to the hypothesis. The manuscript, published in January 2022, describes data from 2013, which may or may not be relevant to the current primary care veterinarian. Certainly the primary drug used for undesirable behavior in this study (acepromazine) is unlikely to be the primary drug used by primary care veterinarians or behavior veterinarians in 2022.


The prevalence of behavior issues described in this study is low compared to other studies, including previous studies by these authors. Likely limiting the criteria to dogs with undesirable behaviors that required medication, and the retrospective use of electronic records where no specific questions or criteria for veterinarians to inquire about behavior, underestimated the overall prevalence of dogs with undesirable behaviors during 2013. Most cases came from one clinic. Manual retrieval of the data from private practice database requires interpretation of the record by the researchers, for example “muzzle for examination because dogs attempts to bite” is categorized as “aggression,” but the underlying emotion is not described. In my opinion this manuscript does little to clarify canine behavior issues or therapeutic approaches. I grant that the introduction and discussion are well referenced and interesting to read.



Now, let’s look at an interesting paper by Bannasch et al. titled  “The effect of inbreeding, body size and morphology on health in dog breeds.” This study examined two databases: a genetic database from a commercial laboratory and the database of a pet insurance company. Inbreeding is basically the breeding of closely genetically related animals or organisms. Most modern dog breeds were developed 100 to 200 years ago, relatively recently in terms of evolution. So in todays’ dog breeds, there are generally common founders within each breed, who may appear multiple time in a multigenerational pedigree, and there is strong selection for common morphology and size within each breed. The coefficient of inbreeding  (COI) is a calculation that describes the probability of inheriting two copies of the same allele from an ancestor that occurs on both sides of the pedigree. So to calculate the COI you ideally need entire multigenerational pedigrees (without the full pedigree you under estimate the COI).

Recently, direct genotype-based methods have been developed as a genetic measure of the inbreeding level. Most purebred dog breeds have an inbreeding level of >0.1; that would be considered extremely high in other nondomestic species. The goal of this project was to examine the relationship between inbreeding, morphology, size, and health; they also choose to examine brachycephalic breeds separately. Median heterozygosity values (H values) were obtained from a large dataset (49,387 dogs representing 227 breeds) from a commercial DNA testing company. Median heterozygosity is the genetic variability within a population. H values were adjusted to provide estimates of inbreeding. Data on breed health came from an insurance company database, so the dogs in the insurance database were not the same dogs as those in the DNA database. They conclude: “Breed average measures of inbreeding, body weight and cumulative insurance data on morbidity were used to identify a relationship between body size, body morphology, inbreeding and health in dogs. Breeds with higher inbreeding levels required greater amounts of veterinary care, as did brachycephalic breeds, and there were significant differences in required health care between FCI breed groupings. We identified a significant effect of both body size and inbreeding on morbidity across dog breeds with larger sized and more inbred breeds receiving more veterinary care throughout their lives.” Simply put the dogs that provided the DNA data were not the dogs in the insurance database.

While I believe the in-breeding data, I am less comfortable with the “breed heath data” and therefore the correlations drawn between in-breeding and breed health. Breed averages (“breed profiles”) can lead to under or over estimates of the impact of in-breeding on health. “Morbidity events” are the number of dogs experiencing a “veterinary care event,” not including preventative care. A  veterinary care event is anything other than preventative care – so in this study a visit to the veterinarian for lacerations from a dog fight is equivalent to an acute asthmatic attack, immune mediated hemolytic anemia, or a fracture from a fall,  an episode of congestive heart failure, hit by car, or gastric foreign body. In my mind, these are all very different events, yet in this paper they are all considered the same – a veterinary care event. If/when data bases of veterinary care are more robust these data could be expanded, currently I view the correlations (except perhaps brachiocephalic breeds) with a touch of skepticism.


A from UC Davis used data from the computerized veterinary hospital records of the University of California-Davis, Veterinary Medical Teaching Hospital. The study goal was to “statistically evaluate the relationships of neutering at different age periods with regard to the occurrence of major joint disorders, and the cancers in German Shepherd Dog.” The authors point out that their facility is both a primary and tertiary care center, with approximately 50,000 patient visits per year. Data from male and female, intact and neutered, German Shepherd dogs (GSD) from 1/1/2000 to 6/30/2014 was extracted. Age at neutering was grouped as < 6 months; 6-12 months; 12-24 months; and 2 to < 9 years; but sometimes to obtain statistical significance they combined two groups (< 12 months referred to as “early neutering”). Although no attempt was made to differentiate dogs presenting for primary care vs tertiary care, dogs for whom the time of neutering was unavailable were not included. Data was available for 1170 GSD, 705 males (245 neutered, 460 intact) and 465 females (293 neutered, 172 intact). For joint disorders, they looked for associations with age at neutering and hip dysplasia, elbow dysplasia, cranial cruciate ligament tear (CCL), and all joint disorders combined. Malignancies investigated were lymphosarcoma, hemangiosarcoma, osteosarcoma, mast cell tumors, and all cancers combined. For female GSD they compared the development of mammary cancer, pyometra, and urinary incontinence between neutered and intact dogs. Data was expressed as percentage of dogs in each group that developed the condition, and interestingly percent is expressed to two decimal points. It is unclear how describing 1 of 24 as 4.17 percent is more informative that 4%. For survival analysis, Cox proportional hazard models were used to assess risk of developing disease among the neutered and intact groups. When combined (< 6 & 6-12 month) there was a significant difference in the development of CCL between early neutered males and females and the intact dogs. There was no apparent relationship between time of neutering and the cancers studied. Not unexpectedly intact and late neutered females were more likely to have mammary cancer. Most of the dogs in this study were not used for breeding, and the sample had proportionally more intact dogs that is expected in the general population. Percentages are always difficult to compare between groups because the percent affected depends on the total number of dogs in the group. I must admit I am always bothered by changing or combining groups to achieve statistical significance, red flags go up to ask if it is biologically significant.  For example, when they combine all joint diseases studied, there is a significant difference between intact male dogs and the combination of male dogs neutered < 6months with those neutered between 6 and 12 months, however the only significant difference when the joint disease data is not combined is for CCL in dogs neutered at < 12 months. According to the data presented 8 of 84 early neutered male dogs developed some evidence of CCL disease. If your dog is one of the 8 dogs that developed CCL issues it is certainly significant to you, but the biological significance is unclear. Compared to the general population of GSD, the database over represents intact dogs of both sexes, which may again influence the results, for example it may be that those early neutered GSD in the general population do not have an increased incidence of CCL compared to intact dogs. So if you have a GSD when do you neuter them? This is a conversation to have with yourself and your veterinarian, considering your life style and your plans for your dog, as well as the relevant literature.


You don’t have to be a researcher to critically review scientific studies. Start with a healthy degree of skepticism, look at the population studied, determine if it represents the population at large, assess what comparisons were made, and seem if it makes sense to you. Next time we will look at case studies and case series, conference abstracts, reviews and unpack some more science.


Craven AJ, Pegram C, Packer RMA, et al. Veterinary drug therapies used for undesirable behaviours in UK dogs under primary veterinary care. PLoS ONE 2022 17(1): e0261139. https://doi.org/10.1371/journal. pone.0261139


Bannasch D, Famula T, Danner J, et al. The effect of inbreeding, body size
and morphology on health in dog breeds. Canine Med Genet 2021 https://doi.org/10.1186/s40575-021-00111-4


Hart BL, Hart LA, Thigpen AP, Willits NH. Neutering of German Shepherd Dogs: associated joint disorders, cancers and urinary incontinence. Vet Med Sci 2016 2:191–199 DOI: 10.1002/vms3.34




Lana Kaiser MD, DVM, born in Buffalo, NY, received a BA in English from SUNY at Buffalo with plans to be a poet. She is a graduate of Michigan State College of Human Medicine, a Board Certified (Human) Internist, a cattle veterinarian, and a Emeritus Professor in the Departments of Physiology and Medicine, in the College of Human Medicine.

Trained as a biomedical researcher with a research focus on cardiovascular pathophysiology and parasitology, she is also interested in scientifically studying the interaction between humans and animals, and has published in both scientific disciplines. A 1995 graduate of the College of Veterinary Medicine, MSU she resides on a farm in Mason, MI where she raises Maine-Anjou and Red Angus cattle.

She has a mobile beef cattle practice, consults for several national agricultural entities, and has written and lectured about animal welfare, animal health, genetic defects, and human-livestock interactions at the state and national levels. She is involved in issues of animal behavior and welfare at the county, state and national level, was a founding member of MSU’s Human-Animal Bond Initiative and coordinator of MVMAs Annual Animal Welfare Conference and Animal Behavior Conference. She has been active in the MVMA Animal Welfare, Legislative, and Food Animal Practivce Committees, AABP Animal Welfare Committee, and Dean Foods Animal Welfare Advisory Council.

She enjoys clicker training her four dogs, her cow horse, and the occasional cow.D