Study animals were identified among all dogs visiting the Small Animal Veterinary Clinic of the University of Veterinary Medicine and Pharmacy in Kosice, Slovak Republic for regular vaccination, parasite treatments and various health complaints. The age of study dogs ranged from 8 to 18 years (mean = 11 years), there were 116 males and 99 females of different breeds, body weights ranged from 2.3 to 55 kg (mean = 19.6 kg) and reproduction state was neutered males (n = 88), neutered females (n = 79); intact males (n = 25), and intact females (n = 23).
Haematology and biochemistry
An integral part of the diagnosis was haematological and biochemical blood tests. The following haematological variables were analysed by IDEXX ProCyte Dx® Hematology Analyzer: HCT, RBC, HGB, MCV, MCH, MCHC, red cell distribution width (RDW), reticulocytes (absolute number and percentage), WBC, neutrophils, lymphocytes, monocytes, eosinophils, and basophils (number and percentage), platelets [number, MPV, platelet distribution width (PDW) and PCT], band neutrophils (when presence suspected) and nRBCs (nucleated red blood cells, when presence suspected). Biochemical blood parameters were analysed by Cobas C 111 analyzer (Roche). These variables included ALT, AST, ALP, pAMS, LIP, Crea, UREA, Glu, Chol, TP, Alb, Ca, P, Mg, NH3, K, Na, Cl. All dogs included in this study were assessed by neurological examination, orthopaedic, radiology, ultrasound and ECG examination, as well as by blood and urine analyses. Eighty-five dogs were excluded from the study because of other medical causes interfering with cognitive decline such as blindness, deafness, diabetes mellitus, cushing syndrome, urinary tract infection, incontinence of urine or faeces, cardiological patients, head trauma and other disease conditions.
Behavioural investigation included observation of geriatric dogs by a veterinary clinician and collection of information provided by pet owners. The investigator was pro-active in asking about behavioural abnormalities to identify even subtle signs that often go unrecognized by pet owners. Data collected by using questionnaire was important for calculation of the final score. The questionnaire also included basic information about dog characteristics and lifestyle variables such as sex, age, weight, reproductive state, dog’s housing and type of diet. The composite scale–CAnine DEmentia Scale (CADES) used in this study was adapted and modified from the questionnaires proposed by Osella, et al.  and Salvin et al. . It contained 17 items distributed into four domains (spatial orientation, social interactions, sleep-awake cycles and house soiling) related to changes in dog’s behaviour. The value of each item corresponded to the frequency of abnormal behaviour. We used a 5-point scale for easy evaluation of behaviour: 0—abnormal behaviour of the dog had never observed; 2—abnormal behaviour of the dog was detected at least once within the last 6 months; 3—abnormal behaviour appeared at least once per month; 4—abnormal behaviour was seen several times per month; 5—abnormal behaviour was observed several times a week. The score from each domain was added up to obtain a final quantitative score that reliably reflected the qualitative evaluation of cognitive decline. We validated CADES as a screening tool for CCDS . Dogs were divided into two subgroups: dogs with no or very mild cognitive impairment (MiCI) and dogs with an advanced cognitive decline. The first group consists of cognitively normal dogs (NA; CADES score 0–7) and dogs with MiCI (CADES score 8–23). The second group consists of dogs with moderate cognitive impairment (MoCI; CADES score 23–44) and dogs with severe cognitive impairment (CADES score higher than 44).
The prevalence of CCDS in three age groups was calculated. All tested dogs were divided into age groups based on estimated life expectancies : short-lived, 6–11 years; medium-lived, 11–13 years and long-lived, >13 years, as proposed by Salvin et al. . The prevalence was calculated by dividing the number of dogs in each tested group (NA, MiCI, MoCI, CD) by the number of individuals examined in each age category.
Statistical analyses were performed with R software . Univariable logistic regression analysis (two-sample Z-test of log odds)  was used to assess relationships between cognitive impairment as dependent variable (moderate and severe CCDS vs normal ageing and mild cognitive impairment) and risk factors such as sex (males vs females), reproductive state (neutered vs entire dogs), food (uncontrolled-scrabs, commercial dry or wet food low quality or mix of different kind of food vs controlled diet—commercial dry or wet food for specific breed, age or life stages—obese, neutered, intact, working dogs), dog’s housing (outside—dogs spending much of their time outside the house vs inside—dogs spending much of their time inside the house or flat) and weight (under 15 kg vs over 15 kg, under 15 kg means smaller or equal than 15 kg in the paper) as independent variables. All age and weight related intervals used in the text are open from left and closed from right, e.g. 11–13 years means greater than 11 and smaller or equal than 13 (except for the first in the range, which is closed from both sides, e.g. 8–11 years means greater or equal than 8 and smaller or equal than 11). The risk was considered to be positive with respect to cognitive impairment when the estimate of odds ratio (OR) was greater than one, and negative when OR was less than one. Additionally, Wald 95 % empirical confidence intervals for OR were calculated. To test correlation between age (in years) and composite CADES score, one-sample Fisher Z-test of zero correlation was used. Additionally, Wald 95 % empirical confidence intervals for Pearson product-moment correlation coefficient were calculated. All of the null hypotheses were tested against two-sided alternatives on significance level α = 0.05. Finally we also used saturated additive multivariable logistic regression model with interaction (of sex and reproductive state) of the form-cognitive impairment ~ sex + reproductive state + diet + housing + weight + (sex: reproductive state), where the log odds were tested similarly as in univariate analyses equivalent to univariable logistic regression model.