Higher burden of neuropsychiatric symptom-like behaviors associated with canine cognitive dysfunction compared to normal aging in the Dog Aging Project

Abstract

Non-cognitive, neuropsychiatric symptoms (NPS) are nearly universal in Alzheimer’s disease (AD), but investigation of their underlying biology is complicated by comparative medicine approaches that incompletely capture spontaneous disease, primarily using transgenic rodent models. The aged companion dog, which spontaneously develops an AD-like disease called canine cognitive dysfunction (CCD), may help fill this translational gap. Using data from the Dog Aging Project with > 10,000 aged dogs (> 8 years old), we identify numerous behaviors in dogs “at-risk” for and with CCD that mirror NPS in humans. Compared to dogs without CCD, our analysis shows that dogs with CCD are less physically active, exhibit fewer previously trained behaviors, demonstrate fewer motivated behaviors, have more daytime sleep, demonstrate more separation anxiety, have altered anxious responses to novelty, have changes in aggressive behaviors, and exhibit lower appetite. Using k-means clustering, we did not find evidence for behavioral sub-phenotypes. Overall, our analysis of a large number of aged dogs suggests clinically significant NPS are associated with CCD and that the companion dog may serve as an important comparative medicine approach to understand these debilitating symptoms across species.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s11357-025-02083-w.

Introduction

While Alzheimer’s disease (AD) is known as a neurodegenerative cause of decreasing cognition in older adults, there are other neuropsychiatric symptoms (NPS) that spontaneously present throughout the disease course. In fact, > 90% of AD patients develop non-cognitive NPS [1], and the appearance of NPS is among the core criteria of the early clinical syndrome [2]. AD-related NPS are broad and include apathy, anxiety, agitation, aggression, appetite changes, dysphoria, and circadian disruptions [3]. NPS are associated with shortened time to institutionalization [4, 5] and reduced quality of life for both patients and caregivers [6–8]. With the exception of brexpiprazole, which was approved for agitation in 2023 [9], there are no FDA-approved medications for any NPS in AD. Instead, off-label pharmacological strategies including antipsychotics are often used but have higher risks for adverse effects, including sedation and even increased mortality [10]. Further, for NPS like depression, the first-line treatments for analogous, primary psychiatric disorders do not lead to benefit in people with dementia [11]. Therefore, it is imperative that new approaches to understanding NPS are undertaken.

Much of the understanding for cognitive symptoms caused by putative AD neuropathology has come from rodent models of AD [12]. However, there are many imperfections with these models that include but are not limited to infrequently modeling age-related components of AD pathogenesis, reliance on familial mutations that may not capture similar pathological processes as spontaneous AD, and lack of overall genetic and environmental heterogeneity [13, 14]. Though NPS-like behaviors have been documented in these murine models [15, 16], the characteristics of these behaviors remain sparse and hard to interpret in non-naturalistic settings [17]. Robust comparative medicine approaches have been central to the development of many new therapeutic strategies, but with the recent failures of rodent models for treatment development for neuropsychiatric disease [18], increasing the tools to study NPS in AD beyond rodents seems likely to lead to better therapeutics.

Fortunately, companion dogs may serve as an excellent opportunity to understand more about NPS in dementia: Unlike mice, aged dogs can spontaneously develop canine cognitive dysfunction (CCD), a syndrome very similar to AD with identifiable amyloid deposits and possibly phosphorylated tau observed in post-mortem brains [19–23]. Unlike transgenic rodents, companion dogs can have a high degree of genetic variability that has been shaped by artificial selection, which can facilitate meaningful forward genetic studies into NPS [24]. Further, dogs live alongside humans and are affected by similar environmental risk factors, receive medical care and veterinary documentation that is comparable to humans, and have a lifespan that facilitates enough time to track the effects of aging while being short enough to provide adequate tissue when the dogs die of natural causes [24–27].

Importantly for NPS in dementia, the relationship between owners and companion dogs with CCD is very similar to the relationship between patients and caregivers in AD. Therefore, identifying clinically significant NPS in dogs may be similar to how these symptoms are identified in human clinical practice and may further allow dog owners to identify CCD more readily. For instance, the most commonly used instrument to detect and measure NPS in humans is the Neuropsychiatric Inventory (NPI), which relies on caregiver report due to the high levels of anosognosia in those with dementia [28, 29]. Similarly, companion dogs’ behavior is observed by their owners daily, and so owners may be able to recognize and report even subtle changes in their dogs’ behavior. Some NPS-like behaviors related to anxiety are even captured by certain veterinarian instruments [30, 31], but NPS-like behaviors related to apathy and aggression are less well-captured in these instruments.

To determine how NPS-like behaviors may manifest in aged companion dogs “at-risk” for and with CCD, we used data from the Dog Aging Project (DAP), a study of normative aging in privately owned companion dogs in the USA [25, 32]. In this large cohort study, we identify many complex NPS-like differences in dogs’ behavior that are associated with CCD and parallel similar varied human NPS that develop during AD.

Methods

The Dog Aging Project

The DAP is a nationwide, prospective cohort study with rolling enrollment that aims to characterize factors related to canine aging through yearly assessments [25]. This open-data, community science approach includes survey data collection, biospecimen collection, and adjunct clinical trials, though only survey data that are germane to this study are further described. This includes all participants in the “DAP Pack.” Starting in 2018, interested companion dog owners agreed to complete a series of yearly questionnaires including the Health and Life Expectancy Survey (HLES) and Canine Social and Learned Behavior (CSLB) survey, both described in more detail below. Owners were recruited via mainstream and social media as well as word of mouth, with enrollment facilitated through the DAP website. Informed consent was obtained from each owner with the ability to withdraw from the study at any time—though the limited dog owner demographic, lifestyle, and anecdotal health information collected in the study met the qualifications for IRB Exempt Status. Specifically, the University of Washington IRB deemed that recruitment of dog owners for the Dog Aging Project and the administration and content of the DAP Health and Life Experience Survey (HLES) are human subject research that qualifies for Category 2 exempt status (IRB ID no. 5988, effective 10/30/2018). No interactions between researchers and privately owned dogs occurred; therefore, IACUC oversight was not required. Data used in this study were obtained from the DAP V1.1 data release on 5/10/2021, representing up to 2 years of study data from participants. At the time of this release, 33,172 dogs were enrolled, with 24,921 having completed at least one HLES and CSLB. To study NPS-like behaviors in dogs at the older end of the lifespan, only dogs aged at least 8 years were included, similar to the population of dogs studied in the validation of the Canine Cognitive Dysfunction Rating Scale (CCDR) (33). This led to 10,027 total dogs in the final analysis. For dogs with more than one assessment (n_all = 15,720, n_old = 5659), only the most recent assessment was included for analysis.

Survey data and data processing

The HLES is an extensive survey capturing numerous aspects of dog health and experience including demographics (age, breed, neuter status, etc.), physical activity, environmental factors, specific behaviors, diet, and health status [33] (https://github.com/dogagingproject/dataRelease/tree/master/Survey_Instruments/HLES). To identify potential NPS-like behaviors included in this analysis, a human psychiatrist (DWF) extracted and categorized terms likely to relate to known NPS domains in human AD (hereafter referred to as “domains”), and these terms were then agreed upon among study authors, which included both human physicians and veterinarians. HLES data underwent extensive cleaning and included (1) removing numeric placeholders for NA values, (2) reversing certain Likert coded variables to consistently reflect similar directionality across the dataset, (3) calculating weekly averages for certain activities coded as frequency and daily duration, (4) combining certain variables of similar behaviors (such as swimming in different water temperatures to a single swimming variable), and (5) recoding values marked < 1 to be 0.5. We estimated lifespan quantiles for each participant based on previous analyses [34], which demonstrated a strong association for median survival time based on weight regardless of whether dogs were pure or mixed breed. These median survival times were used as max expected lifespan, and we assumed a uniform distribution to create a percentile of their expected lifespan from their age. Participant ages above the expected lifespan value were coded as percentiles above the 100th. For example, if a dog lived 20% longer than its expected median survival time, it would be in the 120th percentile of life. To better understand potential NPS-like behaviors, behaviors were aggregated based on expert opinion into numerous categories by the likelihood of belonging to underlying neuropsychiatric domains (“aggregated domains”). If there were missing values, that participant was not included in these aggregate domains. Three variables (“Frequency returns when called off-leash,” “Frequency of other aerobic activity,” “Time interacting with other animals”) were left unsorted, as a high number of missing values precluded grouping of these variables.

The CSLB was adapted from the CCDR [35], with slight modifications [36] including changing the instrument name to avoid mentioning dementia/cognitive dysfunction and changing spelling to reflect American English norms. Briefly, 13 questions are Likert scored and assess the frequency of CCD-related behaviors as observed by the owner. Some questions further prompt the owner to describe changes in these behaviors over a 6-month period. Unchanged from the original CCDR, certain questions had multipliers based on clinical severity, leading to a final possible score range of 16–80. A dog was considered likely to have CCD if their total score was > 50. Similar to other analyses [37–39], a score of 40–50 was also noted for those at “at-risk” or “early-stage” CCD. The CSLB was used to classify dogs’ CCD status, while the HLES was used to derive NPS-like behaviors.

Analysis and statistics

The primary outcomes of interest were how CCD status associated with NPS-like behavior. Two approaches were taken to investigate these relationships: First, each individual survey item that could be related to an NPS-like behavior was treated as its own dependent variable. Second, individual survey items that were likely to be related to distinct NPS-like domains were aggregated into a domain score. To achieve this, each individual score underwent z-score normalization, and the mean of each group of z-scores was calculated to reflect the scores for each domain. Once these terms were generated, a series of linear regression models were calculated. For the individual NPS and domain variables, two models were generated: an unadjusted model and a covariate-adjusted model including expected lifespan percentile, sex, spayed/neutered status, and weight. A Benjamini–Hochberg adjusted p-value was calculated separately for the individual and domain score analyses. Adjusted significance levels were used to scale confidence intervals (CI₉₅) to maintain the overall false-discovery rate control—thus adjusted CI₉₅ represent confidence intervals that correspond to the adjusted p-value.

Principal component analysis (PCA; FactoMineR [40], R version 4.3.2 [41]) was performed on individual NPS-like survey items to assess the underlying variance structure. Some items with a high number of missing values were removed, though results were similar with these items retained and using the subsequent smaller number of included participants. Resultant eigenvalues greater than 1 (Kaiser criterion) and parallel analysis were both used to guide the selection of the number of factors to retain for further analysis; both methods converged on the same number. Exploratory factor analysis (EFA) using the ordinary least squares method (psych package [42], R) was performed to identify latent constructs underlying the data.

To assess whether individual participants formed distinct clusters based on patterns of co-occurring NPS-like domains, unsupervised k-means clustering was performed. Specifically, the fviz_nbclust() function (factoextra [43], R) was used to estimate the optimal number of clusters using both the Elbow method (within-cluster sum of squares) and Average Silhouette Method. As we were primarily interested in possible behavioral phenotypes that may emerge in CCD, we restricted our analyses to participants with CSLB score > 50. In a subsequent analysis, participants with scores between 40 and 50 were also included, yielding similar clustering results. K-means clustering was then performed on the standardized feature set, with cluster visualization performed via the fviz_cluster() function.

Results

The entire dataset included 24,921 unique participants with at least 1 CCD assessment via CSLB. Each individual’s CSLB score was converted into categories for “absent,” “at-risk,” and “present” for CCD. Distributions based on expected lifespan percentile and age are shown in Supplementary Fig. 1. As expected, those with CSLB scores indicating CCD was present, representing 479 participants (1.9% of total cohort), were almost exclusively above the 75^th percentile in relation to their expected lifespan, with the mean expected lifespan percentile being 90th and median being 92^nd. Those in the “at-risk” category, representing 3016 participants (12.1%), had a mean expected lifespan percentile of 64^th and a median of 71^st. For comparison, dogs without CCD had a mean expected lifespan percentile of 43rd and median of 41^st. Similarly by age, those with CCD had a mean age of 14.0 and a median age of 14.1 years old, while those “at risk” had a mean age of 9.8 and a median age of 11.0 years old. Dogs without CCD were younger, with a mean age of 6.1 and median age of 6.5 years old.

When the analysis was restricted to the 10,027 participants older than 8 years, those with CCD (n = 465; 4.6% of older dogs) were in similar expected lifespan percentiles, with a mean in the 92nd percentile and median in the 93^rd percentile. However, when only looking at dogs older than 8 years old, those “at-risk” for CCD (n = 2042; 20.4% of older dogs) had mean and median in the 82nd expected lifespan percentile, which was higher when compared to all the participants in DAP, suggesting some dogs categorized as “at-risk” may be younger than the last quartile of life, which would be very unlikely for a true CCD prodrome. By comparison, dogs without CCD had a mean in the 72nd expected lifespan percentile and a median in the 70th percentile. By age, the group with CCD had a mean of 14.3 years of age and a median of 14.2 years of age, while the group “at-risk” had a mean and median of 12.5 years of age. In contrast, older dogs where CCD was absent had a mean age of 10.9 years and a median age of 10.6 years. Because of these shifts in expected lifespan percentiles in the groups without CCD and those “at-risk” for CCD, we restricted the rest of the analysis to dogs older than 8 years of age.

As NPS in humans are often categorized into behavioral domains based on symptom co-occurrence [44], we performed EFA on survey items designed to capture canine behaviors potentially resembling NPS-like phenotypes. Prior to EFA, we used PCA (PCA; Supplementary Fig. 2A) for dimensionality assessment and to inform factor retention. Based on inspection of eigenvalues and variance explained, we retained 14 factors, with principal components 1–14 accounting for 60.8% of total variance. The resulting EFA model demonstrated a mean item complexity of 1.7 and a df-corrected root mean square of residuals (RMSR) of 0.02, indicating moderate factor complexity and good model fit. Overall, EFA revealed broadly similar item groupings to expert opinion, though several larger domains showed finer subdivisions within the data-driven solution (Supplementary Fig. 2B,C). These findings suggest that expert-assigned domains closely reflect natural covariance structures in the data, and these expert-defined domains were used in subsequent analyses.

Although CCD is typically diagnosed as a single clinical entity, it is likely that multiple underlying etiologies contribute to its manifestation, potentially resulting in distinct clinical subtypes, analogous to patterns observed in human disease [45]. To explore potential subgroup structure, we applied k-means clustering to participants’ NPS-like survey responses. However, evaluation using the Average Silhouette method did not identify a strong clustering solution; no tested cluster number yielded an average silhouette width exceeding 0.2, indicating limited evidence for discrete subgroups and suggesting relative homogeneity within the sample. For completeness, solutions using 2–3 clusters are presented in Supplementary Fig. 3.

To better understand how behaviors change with CCD, we investigated the relationships between CCD status using the CLSB scores and both individual survey items and aggregated behavioral domains. We compared dogs “at-risk” for CCD and those with CCD to dogs with a CSLB score indicating no CCD (reference group) and adjusted for expected lifespan percentile, sex, spayed/neutered status, and weight. As expected, there were a number of behaviors that were associated differently with CCD status (Tables 1 and 2).

Table 1.

Aggregate behavioral domain scores: Based on expert-opinion, individual survey items were z-score normalized and grouped into behavioral domains. Linear regression without covariates (minimal model) and with adjustments for expected lifespan percentile, sex, spayed/neutered status, and weight (covariate-adjusted model) are shown. Unadjusted and Benjamini–Hochberg adjusted p-values are shown. Confidence intervals were scaled to adjusted p-values. p-values in bold are < 0.05

			Minimal Model				Covariate-adjusted model
Variable	CSLB score	Count (n)	Coefficient (β)	Adjusted 95% confidence interval	p-value	Adjusted p-value	Coefficient (β)	Adjusted 95% confidence interval	p-value	Adjusted p-value
Aggression: Familiar	< 40	6174	Ref	-	-	-	Ref	-	-	-
	40–50	1692	0.05	0.01 to 0.09	0.008	0.009	0.04	− 0.02 to 0.10	0.192	0.213
	> 50	380	0.12	0.05 to 0.19	0.001	0.002	0.12	0.01 to 0.23	0.017	0.026
Aggression: Unfamiliar	< 40	6174	Ref	-	-	-	Ref	-	-	-
	40–50	1692	− 0.02	− 0.06 to 0.02	0.461	0.461	0.07	0.00 to 0.14	0.036	0.051
	> 50	380	− 0.16	− 0.23 to − 0.09	< 0.001	< 0.001	− 0.05	− 0.16 to 0.06	0.380	0.380
Appetite	< 40	6174	Ref	-	-	-	Ref	-	-	-
	40–50	1692	− 0.16	− 0.20 to − 0.12	< 0.001	< 0.001	− 0.11	− 0.20 to − 0.02	0.004	0.007
	> 50	380	− 0.27	− 0.34 to − 0.20	< 0.001	< 0.001	− 0.12	− 0.27 to 0.03	0.101	0.126
Engagement	< 40	6174	Ref	-	-	-	Ref	-	-	-
	40–50	1692	− 0.17	− 0.22 to − 0.12	< 0.001	< 0.001	− 0.07	− 0.11 to − 0.03	< 0.001	0.001
	> 50	380	− 0.47	− 0.57 to − 0.37	< 0.001	< 0.001	− 0.25	− 0.33 to − 0.17	< 0.001	< 0.001
Fear	< 40	6174	Ref	-	-	-	Ref	-	-	-
	40–50	1692	0.15	0.11 to 0.19	< 0.001	< 0.001	0.19	0.13 to 0.25	< 0.001	< 0.001
	> 50	380	0.15	0.08 to 0.22	< 0.001	< 0.001	0.09	0.00 to 0.18	0.040	0.053
Human–Dog Interaction	< 40	6174	Ref	-	-	-	Ref	-	-	-
	40–50	1692	− 0.04	− 0.08 to − 0.00	0.031	0.035	− 0.04	− 0.10 to 0.02	0.127	0.150
	> 50	380	− 0.06	− 0.13 to 0.01	0.144	0.152	− 0.05	− 0.15 to 0.05	0.340	0.358
Loss of Training	< 40	6174	Ref	-	-	-	Ref	-	-	-
	40–50	1692	0.18	0.13 to 0.23	< 0.001	< 0.001	0.16	0.11 to 0.21	< 0.001	< 0.001
	> 50	380	0.39	0.30 to 0.48	< 0.001	< 0.001	0.29	0.19 to 0.39	< 0.001	< 0.001
Physical Activity	< 40	6174	Ref	-	-	-	Ref	-	-	-
	40–50	1692	− 0.28	− 0.33 to − 0.23	< 0.001	< 0.001	− 0.22	− 0.29 to − 0.15	< 0.001	< 0.001
	> 50	380	− 0.47	− 0.56 to − 0.38	< 0.001	< 0.001	− 0.32	− 0.42 to − 0.22	< 0.001	< 0.001
Separation	< 40	6174	Ref	-	-	-	Ref	-	-	-
	40–50	1692	0.23	0.18 to 0.28	< 0.001	< 0.001	0.22	0.14 to 0.30	< 0.001	< 0.001
	> 50	380	0.38	0.30 to 0.46	< 0.001	< 0.001	0.33	0.19 to 0.47	< 0.001	< 0.001
Sleep	< 40	6174	Ref	-	-	-	Ref	-	-	-
	40–50	1692	0.24	0.20 to 0.28	< 0.001	< 0.001	0.2	0.10 to 0.30	< 0.001	< 0.001
	> 50	380	0.44	0.36 to 0.52	< 0.001	< 0.001	0.21	0.03 to 0.39	0.009	0.015

Table 2.

Individual behavior scores: Linear regression without covariates (minimal model) and with adjustments for expected lifespan percentile, sex, spayed/neutered status, and weight (covariate-adjusted model) are shown for each individual survey item. Domains were chosen based on expert-opinion. Unadjusted and Benjamini–Hochberg Adjusted p-values are shown. Confidence intervals were scaled to adjusted p-values. p-values in bold are < 0.05

				Minimal Model				Covariate-adjusted model
Domain	Variable	CSLB score	Count (n)	Coefficient (β)	Adjusted 95% confidence interval	p-value	Adjusted p-value	Coefficient (β)	Adjusted 95% confidence interval	p-value	Adjusted p-value
Aggression: Familiar	Aggression when approached by family while eating	< 40	6889	Ref	-	-	-	Ref	-	-	-
		40–50	1822	0.02	− 0.02 to 0.06	0.087	0.099	0	− 0.03 to 0.03	0.903	0.928
		> 50	408	0.05	− 0.02 to 0.12	0.020	0.024	0.05	− 0.01 to 0.11	0.091	0.122
	Aggression against familiar dog while eating	< 40	4254	Ref	-	-	-	Ref	-	-	-
		40–50	1052	0.11	0.07 to 0.15	0.001	0.002	0.08	− 0.03 to 0.19	0.104	0.138
		> 50	234	0.19	0.12 to 0.26	0.004	0.005	0.24	0.04 to 0.44	0.008	0.013
	Aggression against familiar dog while playing	< 40	4424	Ref	-	-	-	Ref	-	-	-
		40–50	1126	0.08	0.04 to 0.12	0.004	0.005	0.11	0.02 to 0.20	0.008	0.013
		> 50	248	0.04	− 0.03 to 0.11	0.466	0.487	0.12	− 0.04 to 0.28	0.112	0.147
	Aggression when food taken away by family	< 40	6853	Ref	-	-	-	Ref	-	-	-
		40–50	1820	0.04	0.00 to 0.08	0.005	0.007	0.03	− 0.01 to 0.07	0.087	0.118
		> 50	409	0.11	0.04 to 0.18	< 0.001	< 0.001	0.1	0.02 to 0.18	0.005	0.010
	Aggression when toys taken away by family	< 40	6892	Ref	-	-	-	Ref	-	-	-
		40–50	1823	0.03	− 0.01 to 0.07	0.068	0.079	0.01	− 0.03 to 0.05	0.616	0.682
		> 50	409	0.08	0.01 to 0.15	0.005	0.007	0.04	− 0.04 to 0.12	0.260	0.325
Aggression: Unfamiliar	Aggression against delivery workers	< 40	6885	Ref	-	-	-	Ref	-	-	-
		40–50	1826	− 0.07	− 0.11 to − 0.03	0.003	0.005	0.04	− 0.04 to 0.12	0.312	0.386
		> 50	408	− 0.35	− 0.43 to − 0.27	< 0.001	< 0.001	− 0.18	− 0.33 to − 0.03	0.009	0.015
	Aggression against unknown dog when on leash	< 40	6867	Ref	-	-	-	Ref	-	-	-
		40–50	1811	0.03	− 0.01 to 0.07	0.243	0.266	0.09	− 0.00 to 0.18	0.027	0.041
		> 50	407	− 0.03	− 0.10 to 0.04	0.556	0.576	0.05	− 0.11 to 0.21	0.519	0.598
	Aggression against unknown human when on leash	< 40	6873	Ref	-	-	-	Ref	-	-	-
		40–50	1816	0.05	0.01 to 0.09	0.019	0.023	0.07	0.01 to 0.13	0.019	0.029
		> 50	407	0	− 0.07 to 0.07	0.965	0.965	0	− 0.11 to 0.11	0.942	0.950
	Aggression when encountering unknown aggressive dog	< 40	6832	Ref	-	-	-	Ref	-	-	-
		40–50	1807	0.02	− 0.02 to 0.06	0.584	0.594	0.14	0.03 to 0.25	0.003	0.006
		> 50	404	− 0.22	− 0.29 to − 0.15	0.000	0.001	− 0.06	− 0.24 to 0.12	0.488	0.568
	Aggression when stranger near yard	< 40	6871	Ref	-	-	-	Ref	-	-	-
		40–50	1824	− 0.05	− 0.09 to − 0.01	0.044	0.052	0.02	− 0.05 to 0.09	0.613	0.682
		> 50	406	− 0.24	− 0.32 to − 0.16	< 0.001	< 0.001	− 0.06	− 0.19 to 0.07	0.364	0.444
Appetite	Appetite Changes	< 40	6906	Ref	-	-	-	Ref	-	-	-
		40–50	1829	− 0.09	− 0.13 to − 0.05	< 0.001	< 0.001	− 0.06	− 0.11 to −0.01	0.011	0.018
		> 50	410	− 0.16	− 0.24 to − 0.08	< 0.001	< 0.001	− 0.08	− 0.16 to 0.00	0.045	0.064
Engagement	Chases birds if given chance	< 40	6897	Ref	-	-	-	Ref	-	-	-
		40–50	1821	− 0.25	− 0.29 to − 0.21	< 0.001	< 0.001	− 0.17	− 0.29 to − 0.05	0.001	0.001
		> 50	405	− 0.61	− 0.69 to − 0.53	< 0.001	< 0.001	− 0.29	− 0.50 to − 0.08	0.001	0.003
	Chases small rodents if given chance	< 40	6865	Ref	-	-	-	Ref	-	-	-
		40–50	1820	− 0.38	− 0.43 to − 0.33	< 0.001	< 0.001	− 0.22	− 0.34 to − 0.10	< 0.001	< 0.001
		> 50	405	− 1.01	− 1.11 to − 0.91	< 0.001	< 0.001	− 0.53	− 0.77 to − 0.29	< 0.001	< 0.001
	Frequency chasing tail	< 40	6852	Ref	-	-	-	Ref	-	-	-
		40–50	1817	0	− 0.04 to 0.04	0.890	0.898	0.02	− 0.02 to 0.06	0.413	0.496
		> 50	405	− 0.07	− 0.14 to 0.00	0.004	0.006	− 0.01	− 0.08 to 0.06	0.723	0.779
	Frequency would escape home if given chance	< 40	6877	Ref	-	-	-	Ref	-	-	-
		40–50	1825	0.03	− 0.01 to 0.07	0.359	0.386	0.06	− 0.03 to 0.15	0.163	0.209
		> 50	406	− 0.09	− 0.16 to − 0.02	0.114	0.128	− 0.04	− 0.20 to 0.12	0.663	0.727
	Excitement level before car ride	< 40	6901	Ref	-	-	-	Ref	-	-	-
		40–50	1828	− 0.15	− 0.19 to − 0.11	< 0.001	< 0.001	0.04	− 0.05 to 0.13	0.366	0.444
		> 50	407	− 0.52	− 0.60 to − 0.44	< 0.001	< 0.001	− 0.32	− 0.50 to − 0.14	< 0.001	< 0.001
	Excitement level before walks	< 40	6905	Ref	-	-	-	Ref	-	-	-
		40–50	1825	− 0.13	− 0.17 to − 0.09	< 0.001	< 0.001	0.03	− 0.05 to 0.11	0.445	0.529
		> 50	409	− 0.56	− 0.65 to − 0.47	< 0.001	< 0.001	− 0.33	− 0.51 to − 0.15	< 0.001	< 0.001
	Frequency playful/puppyish	< 40	6899	Ref	-	-	-	Ref	-	-	-
		40–50	1829	− 0.32	− 0.37 to − 0.27	< 0.001	< 0.001	− 0.21	− 0.31 to − 0.11	< 0.001	< 0.001
		> 50	407	− 0.69	− 0.79 to − 0.59	< 0.001	< 0.001	− 0.48	− 0.66 to − 0.30	< 0.001	< 0.001
	Frequency playing physically active games	< 40	7003	Ref	-	-	-	Ref	-	-	-
		40–50	1848	− 0.56	− 0.61 to − 0.51	< 0.001	< 0.001	− 0.33	− 0.48 to − 0.18	< 0.001	< 0.001
		> 50	416	− 1.15	− 1.25 to − 1.05	< 0.001	< 0.001	− 0.67	− 0.95 to − 0.39	< 0.001	< 0.001
Fear	Fear when groomed by family	< 40	6798	Ref	-	-	-	Ref	-	-	-
		40–50	1786	0.31	0.26 to 0.36	< 0.001	< 0.001	0.32	0.21 to 0.43	< 0.001	< 0.001
		> 50	400	0.49	0.40 to 0.58	< 0.001	< 0.001	0.42	0.24 to 0.60	< 0.001	< 0.001
	Fear in response to loud noises	< 40	6878	Ref	-	-	-	Ref	-	-	-
		40–50	1826	0.05	0.01 to 0.09	0.123	0.137	0.14	0.03 to 0.25	0.008	0.014
		> 50	404	− 0.38	− 0.46 to − 0.30	< 0.001	< 0.001	− 0.24	− 0.44 to − 0.04	0.011	0.018
	Fear nails clipped by family	< 40	6529	Ref	-	-	-	Ref	-	-	-
		40–50	1710	0.33	0.28 to 0.38	< 0.001	< 0.001	0.36	0.22 to 0.50	< 0.001	< 0.001
		> 50	379	0.42	0.34 to 0.50	< 0.001	< 0.001	0.27	0.07 to 0.47	0.003	0.006
	Fear of unknown aggressive dog	< 40	6816	Ref	-	-	-	Ref	-	-	-
		40–50	1793	0.13	0.09 to 0.17	< 0.001	< 0.001	0.15	0.05 to 0.25	0.001	0.002
		> 50	400	0.05	− 0.02 to 0.12	0.420	0.448	− 0.04	− 0.20 to 0.12	0.581	0.656
	Fear when approached by unfamiliar dogs	< 40	6866	Ref	-	-	-	Ref	-	-	-
		40–50	1813	0.11	0.07 to 0.15	< 0.001	< 0.001	0.1	0.01 to 0.19	0.009	0.015
		> 50	404	0.12	0.05 to 0.19	0.027	0.033	− 0.02	− 0.16 to 0.12	0.788	0.824
	Fear when approached by unknown human	< 40	6862	Ref	-	-	-	Ref	-	-	-
		40–50	1822	0.11	0.07 to 0.15	< 0.001	< 0.001	0.15	0.07 to 0.23	< 0.001	< 0.001
		> 50	403	0.17	0.10 to 0.24	< 0.001	< 0.001	0.14	0.01 to 0.27	0.021	0.032
	Fear when touched by unfamiliar human	< 40	6883	Ref	-	-	-	Ref	-	-	-
		40–50	1820	0.14	0.10 to 0.18	< 0.001	< 0.001	0.17	0.09 to 0.25	< 0.001	< 0.001
		> 50	405	0.25	0.17 to 0.33	< 0.001	< 0.001	0.17	0.03 to 0.31	0.005	0.010
	Fear in response to unfamiliar objects	< 40	6873	Ref	-	-	-	Ref	-	-	-
		40–50	1818	0.08	0.04 to 0.12	< 0.001	< 0.001	0.12	0.05 to 0.19	< 0.001	< 0.001
		> 50	406	0.11	0.04 to 0.18	0.002	0.003	0.12	0.01 to 0.23	0.016	0.025
	Fear in response to unfamiliar situations	< 40	6858	Ref	-	-	-	Ref	-	-	-
		40–50	1811	0.24	0.19 to 0.29	< 0.001	< 0.001	0.27	0.16 to 0.38	< 0.001	< 0.001
		> 50	407	0.25	0.18 to 0.32	< 0.001	< 0.001	0.21	0.05 to 0.37	0.002	0.005
Human–Dog Interaction	Tendency to follow family around house	< 40	6915	Ref	-	-	-	Ref	-	-	-
		40–50	1832	0.12	0.08 to 0.16	< 0.001	< 0.001	0.1	0.01 to 0.19	0.016	0.024
		> 50	409	0.22	0.15 to 0.29	< 0.001	< 0.001	0.17	0.02 to 0.32	0.015	0.024
	Tendency to sit close to family when sitting	< 40	6916	Ref	-	-	-	Ref	-	-	-
		40–50	1831	− 0.07	− 0.11 to − 0.03	0.012	0.016	− 0.07	− 0.15 to 0.01	0.075	0.103
		> 50	409	− 0.03	− 0.10 to 0.04	0.567	0.583	− 0.02	− 0.15 to 0.11	0.759	0.808
	Time interacting with human adults	< 40	6954	Ref	-	-	-	Ref	-	-	-
		40–50	1839	− 1.48	− 1.53 to − 1.43	< 0.001	< 0.001	− 1.32	− 2.01 to − 0.63	< 0.001	< 0.001
		> 50	412	− 2.35	− 2.43 to − 2.27	< 0.001	< 0.001	− 1.65	− 2.80 to − 0.50	0.001	0.002
Loss of Training	Barks persistently when alarmed/excited	< 40	6912	Ref	-	-	-	Ref	-	-	-
		40–50	1827	− 0.09	− 0.13 to − 0.05	0.008	0.011	− 0.01	− 0.10 to 0.08	0.891	0.923
		> 50	408	− 0.39	− 0.47 to − 0.31	< 0.001	< 0.001	− 0.23	− 0.42 to − 0.04	0.006	0.011
	Frequency chewing on inappropriate objects	< 40	6891	Ref	-	-	-	Ref	-	-	-
		40–50	1827	0.01	− 0.03 to 0.05	0.448	0.473	0.09	0.03 to 0.15	0.002	0.004
		> 50	406	− 0.05	− 0.12 to 0.02	0.140	0.155	0.02	− 0.08 to 0.12	0.725	0.779
	Frequency defecates when left alone	< 40	6874	Ref	-	-	-	Ref	-	-	-
		40–50	1820	0.24	0.19 to 0.29	< 0.001	< 0.001	0.18	0.11 to 0.25	< 0.001	< 0.001
		> 50	409	0.62	0.50 to 0.74	< 0.001	< 0.001	0.41	0.27 to 0.55	< 0.001	< 0.001
	Frequency hyperactive/restless when left alone	< 40	6862	Ref	-	-	-	Ref	-	-	-
		40–50	1825	0.38	0.33 to 0.43	< 0.001	< 0.001	0.33	0.22 to 0.44	< 0.001	< 0.001
		> 50	403	0.88	0.78 to 0.98	< 0.001	< 0.001	0.77	0.53 to 1.01	< 0.001	< 0.001
	Frequency pulls hard on leash	< 40	6866	Ref	-	-	-	Ref	-	-	-
		40–50	1822	− 0.07	− 0.11 to − 0.03	0.007	0.009	0.05	− 0.03 to 0.13	0.207	0.262
		> 50	406	− 0.27	− 0.35 to − 0.19	< 0.001	< 0.001	− 0.04	− 0.19 to 0.11	0.536	0.611
	Easily distracted by sensory stimuli	< 40	6910	Ref	-	-	-	Ref	-	-	-
		40–50	1826	0.02	− 0.02 to 0.06	0.337	0.366	0.07	− 0.01 to 0.15	0.043	0.062
		> 50	407	− 0.09	− 0.16 to − 0.02	0.068	0.079	− 0.02	− 0.15 to 0.11	0.781	0.824
	Frequency obeys sit command	< 40	6903	Ref	-	-	-	Ref	-	-	-
		40 − 50	1821	0.33	0.28 to 0.38	< 0.001	< 0.001	0.2	0.11 to 0.29	< 0.001	< 0.001
		> 50	402	0.82	0.72 to 0.92	< 0.001	< 0.001	0.57	0.37 to 0.77	< 0.001	< 0.001
	Frequency obeys stay command	< 40	6840	Ref	-	-	-	Ref	-	-	-
		40–50	1808	0.33	0.28 to 0.38	< 0.001	< 0.001	0.22	0.11 to 0.33	< 0.001	< 0.001
		> 50	401	0.76	0.67 to 0.85	< 0.001	< 0.001	0.5	0.30 to 0.70	< 0.001	< 0.001
	Frequency urinates when left alone	< 40	6887	Ref	-	-	-	Ref	-	-	-
		40–50	1823	0.27	0.22 to 0.32	< 0.001	< 0.001	0.19	0.12 to 0.26	< 0.001	< 0.001
		> 50	408	0.7	0.59 to 0.81	< 0.001	< 0.001	0.46	0.31 to 0.61	< 0.001	< 0.001
	Frequency urinates on objects	< 40	6891	Ref	-	-	-	Ref	-	-	-
		40–50	1825	0.1	0.06 to 0.14	< 0.001	< 0.001	0.06	0.02 to 0.10	0.006	0.011
		> 50	407	0.23	0.15 to 0.31	< 0.001	< 0.001	0.15	0.06 to 0.24	< 0.001	< 0.001
Physical Activity	Frequency active/energetic	< 40	6864	Ref	-	-	-	Ref	-	-	-
		40–50	1820	− 0.39	− 0.45 to − 0.33	< 0.001	< 0.001	− 0.28	− 0.38 to − 0.18	< 0.001	< 0.001
		> 50	405	− 0.65	− 0.74 to − 0.56	< 0.001	< 0.001	− 0.46	− 0.64 to − 0.28	< 0.001	< 0.001
	Lifestyle activity level (not active, moderately active, very active)	< 40	7012	Ref	-	-	-	Ref	-	-	-
		40–50	1849	− 0.25	− 0.31 to − 0.19	< 0.001	< 0.001	− 0.2	− 0.27 to − 0.13	< 0.001	< 0.001
		> 50	416	− 0.44	− 0.54 to − 0.34	< 0.001	< 0.001	− 0.31	− 0.42 to − 0.20	< 0.001	< 0.001
	Total Amount of time spent active	< 40	7012	Ref	-	-	-	Ref	-	-	-
		40–50	1849	− 0.3	− 0.35 to − 0.25	< 0.001	< 0.001	− 0.3	− 0.44 to − 0.16	< 0.001	< 0.001
		> 50	416	− 0.39	− 0.47 to − 0.31	< 0.001	< 0.001	− 0.32	− 0.55 to − 0.09	0.002	0.004
	Intensity level when active (low, moderate, vigorous)	< 40	7520	Ref	-	-	-	Ref	-	-	-
		40–50	2042	− 0.17	− 0.22 to − 0.12	< 0.001	< 0.001	− 0.1	− 0.16 to − 0.04	< 0.001	< 0.001
		> 50	465	− 0.31	− 0.40 to − 0.22	< 0.001	< 0.001	− 0.15	− 0.25 to − 0.05	< 0.001	0.001
	Time spent walking on-leash	< 40	6403	Ref	-	-	-	Ref	-	-	-
		40–50	1646	− 0.76	− 0.81 to − 0.71	< 0.001	< 0.001	− 0.61	− 0.94 to − 0.28	< 0.001	< 0.001
		> 50	350	− 1.3	− 1.38 to − 1.22	< 0.001	< 0.001	− 0.73	− 1.27 to − 0.19	0.003	0.005
	Time walking off-leash	< 40	2622	Ref	-	-	-	Ref	-	-	-
		40–50	634	− 1.69	− 1.73 to − 1.65	< 0.001	< 0.001	− 1.37	− 2.47 to − 0.27	0.006	0.011
		> 50	124	− 1.71	− 1.78 to − 1.64	0.017	0.022	− 0.7	− 2.62 to 1.22	0.461	0.542
Separation	Frequency barks/whines when left alone	< 40	6887	Ref	-	-	-	Ref	-	-	-
		40–50	1823	0.2	0.16 to 0.24	< 0.001	< 0.001	0.21	0.11 to 0.31	< 0.001	< 0.001
		> 50	407	0.34	0.26 to 0.42	< 0.001	< 0.001	0.36	0.17 to 0.55	< 0.001	< 0.001
	Frequency of restlessness/agitation when left alone	< 40	6878	Ref	-	-	-	Ref	-	-	-
		40–50	1823	0.28	0.23 to 0.33	< 0.001	< 0.001	0.24	0.15 to 0.33	< 0.001	< 0.001
		> 50	406	0.7	0.59 to 0.81	< 0.001	< 0.001	0.62	0.43 to 0.81	< 0.001	< 0.001
	Frequency scratches at door when left alone	< 40	6888	Ref	-	-	-	Ref	-	-	-
		40–50	1822	0.09	0.05 to 0.13	< 0.001	< 0.001	0.11	0.05 to 0.17	< 0.001	< 0.001
		> 50	406	0.07	0.00 to 0.14	0.040	0.047	0.1	− 0.00 to 0.20	0.035	0.051
Sleep	Estimated hours asleep in daytime	< 40	6271	Ref	-	-	-	Ref	-	-	-
		40–50	1728	0.95	0.90 to 1.00	< 0.001	< 0.001	0.88	0.41 to 1.35	< 0.001	< 0.001
		> 50	387	1.81	1.73 to 1.89	< 0.001	< 0.001	0.92	0.16 to 1.68	0.008	0.013
	Estimated hours asleep in nighttime	< 40	6797	Ref	-	-	-	Ref	-	-	-
		40–50	1794	0.14	0.10 to 0.18	< 0.001	< 0.001	0.01	− 0.10 to 0.12	0.927	0.943
		> 50	401	0.17	0.10 to 0.24	0.016	0.021	0.01	− 0.19 to 0.21	0.959	0.959
	Total time asleep	< 40	6215	Ref	-	-	-	Ref	-	-	-
		40–50	1704	1.08	1.03 to 1.13	< 0.001	< 0.001	0.9	0.40 to 1.40	< 0.001	< 0.001
		> 50	383	1.98	1.90 to 2.06	< 0.001	< 0.001	0.97	0.16 to 1.78	0.009	0.015
Uncategorized	Frequency returns when called off-leash	< 40	2916	Ref	-	-	-	Ref	-	-	-
		40–50	710	0.17	0.13 to 0.21	< 0.001	< 0.001	0.1	− 0.01 to 0.21	0.058	0.081
		> 50	135	0.45	0.37 to 0.53	< 0.001	< 0.001	0.42	0.18 to 0.66	< 0.001	< 0.001
	Frequency of other aerobic activity	< 40	2738	Ref	-	-	-	Ref	-	-	-
		40–50	605	− 0.69	− 0.73 to − 0.65	< 0.001	< 0.001	− 0.82	− 1.36 to − 0.28	< 0.001	0.001
		> 50	91	− 0.78	− 0.85 to − 0.71	0.025	0.031	− 0.77	− 1.80 to 0.26	0.121	0.157
	Time interacting with other animals	< 40	6954	Ref	-	-	-	Ref	-	-	-
		40–50	1839	− 1.19	− 1.23 to − 1.15	< 0.001	< 0.001	− 0.67	− 1.38 to 0.04	0.046	0.065
		> 50	412	− 2.5	− 2.58 to − 2.42	< 0.001	< 0.001	− 1.93	− 3.32 to − 0.54	0.002	0.004
	Weight change in last year	< 40	6886	Ref	-	-	-	Ref	-	-	-
		40–50	1822	− 0.05	− 0.09 to − 0.01	< 0.001	< 0.001	− 0.05	− 0.08 to − 0.02	< 0.001	0.001
		> 50	407	− 0.17	− 0.25 to − 0.09	< 0.001	< 0.001	− 0.11	− 0.17 to − 0.05	< 0.001	< 0.001

Similar to previous reports [46, 47], the physical activity domain was lower for dogs with CCD and those “at-risk” for CCD compared to dogs without CCD in both the unadjusted and adjusted models. In both models, nearly every specific physical activity item was also lower for those “at-risk” for and with CCD, including frequency of being energetic, lifestyle activity level, total time spent active, intensity level when active, and time spent walking on leash. Only dogs “at-risk” for CCD had significantly less time walking off-leash.

As anticipated, dogs “at-risk” for and with CCD had greater loss of training in both the unadjusted and adjusted models. In both models, there were losses of training behaviors related to defecation when left alone, urinating when left alone, hyperactivity/restlessness when left alone, obeying “sit” commands, obeying “stay” commands, and urinating on objects. In the unadjusted models, dogs “at-risk” for and with CCD both had lower scores in barking persistently when alarmed/excited and pulling hard on their leash, but only dogs with CCD had significantly less barking when alarmed/excited in adjusted models. Only dogs “at-risk” for CCD showed a significant association with greater chewing on inappropriate objects in adjusted but not unadjusted models. While the dogs “at-risk” for CCD showed a trend towards being more distracted by sensory stimuli, this was not significant after adjusting for multiple comparisons (p_adj = 0.06). It is noted that for some behaviors like barking when alarmed/excited and pulling hard on the leash, the direction of change may be more reflective of reduced motivation or drive than loss of training, with pulling hard on leash clustering with some of the engagement behaviors in our EFA.

Dogs “at-risk” for and with CCD both had lower scores in the engagement domain, aggregating motivated behaviors, in both the unadjusted and adjusted models. In both models, there were lower scores in engaged activities related to chasing birds, chasing small rodents, frequency of being playful, and playing physically active games. While dogs “at-risk” for and with CCD had less excitement before car rides and less excitement before walks in unadjusted models, only dogs with CCD had significantly less excitement before car rides and before walks in adjusted models. While dogs with CCD had less activity related to chasing their tails in unadjusted models, neither group had significantly less tail chasing in adjusted models. The frequency of escaping home or trying to escape home was not statistically different in either model.

Sleep disruptions were common in dogs “at-risk” for and with CCD, with more sleep being observed in both the unadjusted and adjusted models. While total hours asleep per day was significantly greater for dogs “at-risk” for and with CCD, this was largely driven by owners’ estimates of their dogs sleeping more during the daytime, as there were no significant changes noted during nighttime.

The frequency of abnormal behaviors upon separation and in response to other fears was split into two domains and displayed different patterns. While both dogs “at-risk” for and with CCD displayed greater distressed behavior upon separation in both unadjusted and adjusted models, only dogs “at-risk” for CCD showed significantly more fear behaviors in both models. By contrast, dogs with CCD displayed significantly more fear behaviors in the unadjusted model but only a trend towards more fear in adjusted models after multiple comparisons were taken into consideration (p_adj = 0.05). For separation on the individual behavior level, both dogs “at-risk” for and with CCD showed greater barking/whining and restlessness/agitation when left alone. However, only dogs “at-risk” for CCD had significantly greater scratching at the door when left alone, though dogs with CCD trended towards this result (p_adj = 0.05). Both dogs “at-risk” for and with CCD showed greater fear-related behaviors, including fear when groomed by family, fear with nails clipped by family, fear when touched by an unknown, fear when approached by an unknown human, fear when presented with unfamiliar objects, and fear when in unfamiliar situations. Interestingly, dogs “at-risk” for CCD showed more fear with loud noises, while dogs with CCD showed less fear with loud noises. Similarly, dogs “at-risk” for CCD showed more fear when interacting with an unknown aggressive dog and when approached by an unfamiliar dog, while dogs with CCD did not have significant differences in these behaviors, and the degree of non-significant difference was often opposite of dogs “at-risk” for CCD.

Changes in aggression were slight and nuanced among dogs “at-risk” for and with CCD. Because aggression can be multimodal [48], aggressive behaviors were split into responses to familiar and unfamiliar stimuli. For familiar aggression, both dogs “at-risk” for and with CCD showed small but significantly greater scores in aggressive behaviors in the unadjusted models, but only dogs with CCD showed significantly more familiar aggression in the adjusted models. By contrast, dogs with CCD showed less aggression against unfamiliar targets in the unadjusted models, while dogs “at-risk” for CCD showed significantly more unfamiliar aggression before correction for multiple comparisons but not after. On the individual behavior level, dogs with CCD but not dogs “at-risk” for CCD showed more aggression against other familiar dogs while eating and when their food was taken away by a family member. Though only dogs “at-risk” for CCD showed significantly more aggression against familiar dogs while playing, the effect size for dogs with CCD was similar but non-significant. Neither group showed differences in aggression when approached by family or when family took their toys away. Dogs “at-risk” for CCD but not with CCD showed more aggression against aggressive dogs, against unknown dogs when on leash, and against unknown humans when on-leash. Conversely, dogs with CCD showed significantly less aggression against delivery workers while dogs “at-risk” for CCD did not. Neither group showed significant differences in aggression when a stranger was near their yard.

Appetite as estimated by owners was found to be reduced in both dogs “at-risk” for and with CCD in unadjusted models, but only dogs “at-risk” for CCD had significantly reduced appetite in adjusted models, though again the effect size was similar for dogs with CCD and only missed the threshold for significance after adjustment for multiple comparisons (p_adj = 0.06). Similarly, owners reported that both dogs “at-risk” for and with CCD had reductions in weight over the last year.

Despite differences in these other behaviors, only dogs “at-risk” for CCD showed a significantly lower score in human–dog interactions in unadjusted models but not in adjusted models. However, the individual behaviors in this domain were more complex. Both dogs “at-risk” for and with CCD showed fewer hours per day interacting with humans but conversely had a higher frequency of following humans around the house. Neither dogs “at-risk” for nor dogs with CCD tended to have statistically different frequencies of being in physical contact with humans while sitting.

Discussion

In this analysis of 10,027 dogs that were 8 years or older, those “at-risk” for or with CCD displayed significantly more NPS-like behaviors. On the population level, these behavioral patterns include less physical activity, more amotivation/apathy, more daytime sleeping, and reduced appetite, with more nuanced differences in separation anxiety, fear, and aggression. Despite these clear differences in behavior, we were unable to detect an underlying clustering structure among the data comprising CCD dog behavior grouped into the aggregated behavioral domains, which, if found, may have suggested certain sub-phenotypes or syndromes related to CCD.

Despite good evidence for cognitive changes [49–51] and AD-like neuropathology with CCD [19–23], the clinical diagnosis ultimately relies on the emergence of a constellation of behaviors in older dogs that are not better explained by other disease processes [45]. Canonically, these behaviors are often captured by the acronym DISHA: Disorientation, social Interaction changes, Sleep–wake cycle abnormalities, House soiling, Activity alterations, and sometimes Anxiety [52–54]. We observed changes in many of these behaviors in our population, and additional behaviors related to reduced motivation may be just as salient, similarly noted by other groups [30, 49, 55, 56]. In addition, some behaviors such as sleep, aggression, and anxiety may have more nuanced and specific patterns in CCD, though additional investigations would be needed to determine this.

As reported previously [46, 47], measures of physical activity were consistently lower in dogs “at-risk” for and with CCD. This was observed both in unadjusted and adjusted models where age represented by percentile of expected lifespan was taken into account, suggesting that these differences are not solely due to normal aging. These measures not only captured fewer hours being active but also differences in the intensity of activity and a less active lifestyle. These results likely reflect decreased activity in more familiar or routine environments, as some past experiments have shown increased wandering and locomotor activity in dogs with CCD in the open field test while also showing reduced goal-directed behaviors such as approaching the door [57]. In humans, there are reciprocal relationships between motor activity and cognitive decline, whereby subtle reduction in measures like gait speed can be early signs of cognitive decline and progression to dementia [58], while increased cardiovascular fitness can also help prevent dementia [59]. Similarly, we might speculate that there are similar relationships for dogs “at-risk” for and with CCD, and paradigms that use actigraphy [60, 61] and interventional trials promoting activity in mid-life [62] may help clarify this relationship.

In AD and other human neurodegenerative dementias, apathy without dysphoria emerges as one of the most prevalent NPS and increases as the disease progresses [63, 64]. Though some aspects of apathy can only be inferred in canine populations, our data suggest there are numerous places where apathy-related behaviors may emerge early in CCD and last throughout the course of the disease. In humans, the importance of recognizing apathy early is increasingly being realized, as it is often prominent in mild behavioral impairment, an analogous prodrome to mild cognitive impairment whereby NPS present before clear cognitive deficits [65, 66]. In our dataset, we found that dogs “at-risk” for and with CCD showed less engagement with their environment, less excitement before activities, less time interacting with humans and other animals, and lower appetites compared to dogs without CCD. While changes in human–dog interactions as an aggregated domain were minimal, our findings may represent that dogs with CCD have reduced motivations for social interactions but more separation anxiety when not with familiar humans—as suggested by the findings of less time interacting with humans but more time following humans around the house. Other groups have also noted reductions in motivated behaviors with CCD including less social interaction [30, 55, 56] and smaller durations engaging in the impossible task, which may represent either early giving up or reduced motivation [49]. It may be possible that specific increases in apathy will serve as important markers of CCD, and understanding more about how apathy emerges as part of CCD may inform new understandings of how the underlying corticostriatal circuitry [67] changes with neurodegeneration across species.

Despite anxiety often being considered a core symptom of CCD, nuances in the behavioral manifestations of anxiety were noted in this study. When it came to being separated or left alone, dogs “at-risk” for and with CCD showed more anxious behaviors consistently compared to dogs without CCD. Similarly, when a situation was unfamiliar or novel, such as interacting with an unfamiliar human, both groups showed greater fear behavior compared to dogs without CCD. Even for some activities that required the dog being in vulnerable positions despite being around a familiar human, like grooming and nail clipping, greater fear responses were common, though how much of this was a component of memory impairment is unclear. By contrast, there were some situations, such as being confronted by an unfamiliar dog and human, confronted by an unfamiliar object, or showing a heightened startle response to loud noises, where those “at-risk” for CCD showed heightened fear responses but those with CCD were either less reactive than or similar to dogs without CCD. One could speculate that this represents a decrease in the ability to detect a threat in dogs with severe CCD, but these differences may also be due to the progression of apathy throughout the disease course. Other groups have similarly noted increases in anxiety with CCD [38, 39, 56], though some groups suggested higher anxiety with mild CCD [39] while others found higher anxiety in more severe CCD [56]. For one of the most commonly used CCD screening tools, the Canine Dementia Scale (CADES), anxiety was removed from the final version because it did not reflect the severity of the disease [30]. In one study looking at anxiety as part of normal dog aging, they found that non-social anxiety increased with age while social anxiety did not [68]. Further investigations into how anxiety changes throughout the disease course and how it relates to neuroanatomical and neurochemical changes will be important to understanding this prominent NPS across species.

Agitation and aggression are common in AD, though whether they are a consequence of disinhibition, due to an increased drive for aggressive acts, or responses to increasing difficulties in meeting one’s needs remains unclear [69]. Similar to the complexities in understanding aggression in AD, dogs “at-risk” for and with CCD similarly showed a nuanced pattern of aggression. Mirroring the needs-driven factors of aggression in humans, dogs with CCD but not those “at-risk” for CCD tended to show more aggression when their perceived access to food changed. In contrast, only dogs “at-risk” for CCD tended to have slightly greater aggression against unknown aggressive and unfamiliar dogs, unfamiliar humans, and familiar dogs while playing, which may suggest a change in innate aggressive drives or disinhibition. With dogs at more severe stages of CCD, it is possible that diminished overall motivation keeps these aggressive behaviors from persisting. Another facet that makes studying aggression in companion dogs challenging is that aggression can be a reason for abandonment or euthanasia [48], and so dogs with aggressive tendencies at baseline and ones that develop clinically significant aggression may be underreported in this cohort. Other groups have noticed greater aggression with dogs “at-risk” for and with CCD [56, 70]. It is further interesting to note that studies of aggression throughout the life cycle in dogs tend to show a peak in aggression in middle age but a decrease in more senior years [68, 71]. As agitation and aggression are some of the hardest to treat and most disruptive NPS in human dementia, understanding the manifestations of aggressive behaviors in CCD may lead to novel treatments of these behaviors for both species.

Circadian rhythm disruption and sleep changes are prevalent with AD [72], and we similarly saw greater overall sleep, especially in the day. While this predominance of daytime sleeping may represent more subtle circadian rhythm dysfunction, it is also very possible that these changes are due to the more subjective nature of these sleep time estimates, which could have precluded recognition of sleep time changes at night. Though disruption of the sleep/wake cycle is a core criterion for DISHA, evidence for changes in these patterns is mostly from owner reports. For instance, in a sample of 94 dogs > 8 years old with and without CCD, dogs with CCD were reported to have more sleeping during the day and restlessness at night as well as greater sleep overall [56]. Similarly, a group tracking domains of dysfunction in CCD noted that sleep–wake alterations were reported early in CCD progression [39], though another group studying similar domains found changes in sleep/wake to be the least clearly related to memory and learning [73]. Very few polysomnography studies specifically studying dogs with CCD have been done, but a notable study done during 2 h afternoon naps noted dogs with CCD spend less time in rapid eye movement (REM) and non-REM (NREM) sleep stages [74]. Overall sleep patterns in normal dog aging can be complex but tend towards increased fragmentation with increased NREM sleep during the day and reduced at night as well as an overall lesser amount of time in REM sleep [75]. Additional polysomnography studies in aged dogs with and without CCD may reveal specific, non-age-related changes that could further detail the CCD phenotype.

Human dementia is defined by a decline in cognitive function that leads to impairment in one’s abilities to take care of themselves, but dementia can have multiple etiologies [45]. Neuropathologically, some of these etiologies are more consistent with the emergence of specific behavioral paradigms, such as the visual hallucinations, REM behavior disorder, parkinsonisms, and cognitive fluctuations in Lewy body disorders and disinhibition, apathy, hyperphagia, and stereotypies in frontotemporal dementia. While CCD is thought to reflect processes most similar to AD neuropathologically, some studies have shown that even AD is not a monolith and may have separate subsyndromes [76, 77]. Therefore, we attempted a k-means clustering analysis to determine if there are behavioral phenotype clusters that might represent these subsyndromes, but we were unable to detect any clusters in the data. While this could represent homogeneity in CCD pathogenesis, it is more likely a limitation of the available data, and future studies may be able to better delineate clinically relevant subtypes.

While the large number of participants is a notable strength of this study, there are a few limitations that are important to note. Certain items on the CSLB—which was the instrument used to categorize dogs as having CCD, “at-risk” for CCD, or without CCD—strongly overlap with the wording of questions on the HLES. Therefore, one limitation is that the findings related to these similar behaviors may be tautological. Specifically, the CSLB queries about (1) pacing without purpose or direction, which is similar to the HLES asking about restlessness when left alone (part of loss of training domain), (2) failure to recognize people or pets, similar to the time spent interacting with humans and other animals in the HLES (human–dog interaction domain), (3) walking away while or avoiding being petted, similar to fear when being approached or touched by an unfamiliar human in the HLES (fear domain), (4) change in defecation in previously kept clean areas over 6 months, similar to HLES frequency of defecation when left alone (loss of training domain), and (5) reduction in activity over 6 months, which may be similar to time spent active in the HLES (physical activity domain). Other items in the CSLB are unique and query behaviors that may be more clearly impacted by cognitive deficits, such as getting stuck behind objects, walking into doors or walls, or ability to find food dropped on the floor. Finally, the CSLB touches on what might be considered severe apathy-like behavior, as it asks about the frequency of dogs staring blankly at the walls or floors. Despite these few specific overlapping items which suggest certain NPS-like behavior changes, the value of the current study is in characterizing many additional behaviors—including some behaviors that may be much more subtle than in the CSLB but still related to similar behavioral domains. Overall, the few overlapping items were singular parts of the composite scores for each of the studied NPS-like domains, and, while contributing to the strength of the association, were infrequent enough to where they did not solely drive these associations.

Other limitations should also be considered. First, while using the CSLB to detect CCD has been well validated, it has been shown to be less sensitive to early signs of CCD and may underreport those “at-risk” compared to other instruments like the CADES [39]. Even the nomenclature of “at-risk” used in this paper may misrepresent a milder stage of CCD where neuropathological or structural brain changes are evident. Therefore, our data may miss subtle early signs of the disease, and changes found in our CCD group likely represent participants with more severe CCD. It is also worth repeating that a diagnosis of CCD is dependent on a full veterinary exam and exclusion of other disease mimics, which was not feasible for this large dataset. In addition, while there is good evidence for cognitive dysfunction correlating with our scale [51], the lack of direct cognitive and veterinary assessments specifically to define CCD is a limitation. Further, though the cohort is prospective, this analysis was cross-sectional due to sparse follow-up in the early stages of this emerging dataset. Therefore, we were unable to determine if these behaviors increased or decreased in time as CSLB scores changed for each subject longitudinally. For some measures like overall activity, time asleep, and appetite, the subjective nature of the reporting may lead to some imprecision, suggesting an opportunity for follow-up with newer actigraphy and EEG technologies. The design of this study was to understand dog aging holistically, and so many of our measures were not designed specifically to evaluate NPS-like behaviors that could be more indicative of CCD compared to other aging-related diseases. While this is a challenge in human neuropsychiatry as well, this also represents an opportunity to design instruments to measure the severity of CCD-related NPS, which may yield tools for earlier diagnosis and help track progression or improvements in these NPS in future canine research. Finally, it has been noted that participants in DAP are broadly healthier than the general population, which may reduce the generalizability of these results.

Overall, this large study confirms that NPS-like behaviors are more common in dogs “at-risk” for and with CCD than in otherwise healthy, aged dogs. The emergence of significant behaviors related to amotivation and apathy may be a defining feature of CCD that is alluded to in other work but could be more formalized in future assessments. Early changes in apathy may be harbingers of advancing CCD, and so noting these behavioral changes clinically could be valuable for veterinarians. Our findings show that privately owned companion dogs may be useful to understand the non-cognitive behavioral components related to neurodegenerative dementias. Our findings will hopefully inspire others to consider companion dogs as important modes of investigation to understand neurodegenerative dementias as well as serve as an opportunity to understand NPS with a comparative medicine approach.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contribution

All authors contributed to writing—review and editing. D.W.F.: conceptualization, methodology, formal analysis, data curation, writing—original draft; J.R.F.: formal analysis; S.R.U.: conceptualization; N.A.T.: data curation; K.F.K.: formal analysis; M.D.: conceptualization, supervision, methodology, writing—original draft.

Dog Aging Project Consortium authors

Joshua M. Akey¹, Elhanan Borenstein^2,3,4, Marta G. Castelhano^5,6, Amanda E. Coleman⁷, Kate E. Creevey⁸, Matthew D. Dunbar⁹, Virginia R. Fajt¹⁰, Jessica Hoffman¹¹, Erica C. Jonlin^12,13, Matt Kaeberlein¹², Elinor K. Karlsson^14,15, Kathleen F. Kerr¹⁶, Jing Ma¹⁷, Evan L. MacLean¹⁸, Stephanie McGrath¹⁹, Natasha Olby²⁰, Daniel E. L. Promislow²¹, May J. Reed²², Audrey Ruple²³, Stephen M. Schwartz^24,25, Sandi Shrager²⁶, Noah Snyder-Mackler^27,28,29, M. Katherine Tolbert⁸.

¹Lewis-Sigler Institute for Integrative Genomics, Princeton University, Princeton, NJ, USA.

²Department of Clinical Microbiology and Immunology, Sackler Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel.

³Blavatnik School of Computer Science, Tel Aviv University, Tel Aviv, Israel

⁴Santa Fe Institute, Santa Fe, NM, USA

⁵Cornell Veterinary Biobank, College of Veterinary Medicine, Cornell University, Ithaca, NY, USA

⁶Department of Clinical Sciences, College of Veterinary Medicine Cornell University, Ithaca, NY, USA

⁷Department of Small Animal Medicine and Surgery, College of Veterinary Medicine, University of Georgia, Athens, GA, USA.

⁸Department of Small Animal Clinical Sciences, Texas A&M University College of Veterinary Medicine & Biomedical Sciences, College Station, TX, USA.

⁹Center for Studies in Demography and Ecology, College of Arts and Sciences, University of Washington, Seattle, WA, USA.

¹⁰Department of Veterinary Physiology and Pharmacology, College of Veterinary Medicine & Biomedical Sciences, Texas A&M University, College Station, TX, USA.

¹¹Department of Biological Sciences, Augusta University, Augusta, GA, USA.

¹²Department of Laboratory Medicine and Pathology, University of Washington School of Medicine, Seattle, WA, USA.

¹³Institute for Stem Cell and Regenerative Medicine, University of Washington, Seattle, WA, USA.

¹⁴Bioinformatics and Integrative Biology, Chan Medical School, University of Massachusetts, WorcesterUSA.

¹⁵Broad Institute of MIT and Harvard, CambridgeUSA.

¹⁶Department of Biostatistics, University of Washington, Seattle, WA, USA.

¹⁷Division of Public Health Sciences, Fred Hutchinson Cancer Research Center, Seattle, WA, USA.

¹⁸College of Veterinary Medicine, University of Arizona, Tucson, AZ, USA.

¹⁹Department of Clinical Sciences, College of Veterinary Medicine and Biomedical Sciences, Colorado State University, Fort Collins, CO, USA.

²⁰Department of Clinical Sciences, College of Veterinary Medicine, North Carolina State University, Raleigh, NC, USA.

²¹Jean Mayer USDA Human Nutrition Research Center on Aging, Tufts University, BostonUSA.

²²Department of Medicine, Division of Gerontology and Geriatric Medicine, University of Washington School of Medicine, Seattle, WA, USA.

²³Department of Population Health Sciences, Virginia-Maryland College of Veterinary Medicine, Virginia Tech, Blacksburg, VA, USA.

²⁴Department of Epidemiology, University of Washington, Seattle, WA, USA.

²⁵Epidemiology Program, Fred Hutchinson Cancer Research Center, Seattle, WA, USA.

²⁶Collaborative Health Studies Coordinating Center, Department of Biostatistics, School of Public Health, University of Washington, Seattle, WA, USA.

²⁷School of Life Sciences, Arizona State University, Tempe, AZ, USA.

²⁸Center for Evolution and Medicine, Arizona State University, Tempe, AZ, USA.

²⁹School for Human Evolution and Social Change, Arizona State University, Tempe, AZ, USA.

Funding

This research is based on publicly available data collected by the Dog Aging Project, under U19 grant AG057377 (PI: Daniel Promislow) from the National Institute on Aging, a part of the National Institutes of Health, and by additional grants and private donations, including generous support from the Glenn Foundation for Medical Research, the Tiny Foundation Fund at Myriad Canada, and the WoodNext Foundation. These data are housed on the Terra platform at the Broad Institute of MIT and Harvard. The Brain Health Study is supported by National Institute on Aging grants R24AG073137 (PIs: Keene and McGrath) and U19AG057377 (PI: Promislow).This research is also supported by the UW Medicine Department of Psychiatry and Behavioral Sciences’ Clinician Scientist Training Program (DWF), the NACC/AA New Investigator Award Program (DWF), and the NIH/NIA (T32AG052354, DWF; AG062514, MD; AG067193, MD).

Data Availability

The curated data from the Dog Aging Project that support these findings are openly available after appropriate request at https://dogagingproject.org/data-access. The analyzed data supporting the findings of this study are available on request from the corresponding author.

Declarations

Ethics approval

Nothing to declare.

Competing interests

The authors declare no competing interests.