Associations between osteoarthritis and duration and quality of night-time rest in dogs

Abstract

Human patients with chronic pain from osteoarthritis often report impaired sleep, but it is not yet known if sleep is also impaired in dogs with osteoarthritis. This study aimed to compare the night-time sleep behaviour of osteoarthritic (N=20) and healthy control (N=21) dogs over a 28-day period, using an actigraphic device (the FitBark activity monitor) and an owner questionnaire designed to measure sleep quality (the SNoRE). Actigraphic data were aggregated to estimate the time each dog spent resting each night, and questionnaires were completed every 7 days. Data were analysed using robust mixed-effects linear regression. The presence of clinical signs of osteoarthritis had a significant effect on actigraphic recordings, with osteoarthritic dogs spending lower proportions of the night period resting (and therefore higher proportions of the night period active) compared to control dogs (z=2.21; P=0.0268). However, there was no significant difference between the SNoRE scores of osteoarthritic and control dogs (z=-1.01, p=0.312). The actigraphic findings of this study suggest that dogs with osteoarthritis may experience impaired sleep, which could have important welfare implications and merits further study.

Full text

Applied Animal Behaviour Science 253 (2022) 105661
Available online 27 May 2022
0168-1591/© 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Associations between osteoarthritis and duration and quality of night-time
rest in dogs
Melissa Smith, Michael Mendl
*
, Joanna C. Murrell
1
Bristol Veterinary School, University of Bristol, Langford House, Langford BS40 5DU, UK
ARTICLE INFO
Keywords:
Actigraphy
Chronic pain
Dog
Osteoarthritis
Sleep
ABSTRACT
Human patients with chronic pain from osteoarthritis often report impaired sleep, but it is not yet known if sleep
is also impaired in dogs with osteoarthritis. This study aimed to compare the night-time sleep behaviour of
osteoarthritic (N =20) and healthy control (N =21) dogs over a 28-day period, using an actigraphic device (the
FitBark activity monitor) and an owner questionnaire designed to measure sleep quality (the SNoRE). Acti-
graphic data were aggregated to estimate the time each dog spent resting each night, and questionnaires were
completed every 7 days. Data were analysed using robust mixed-effects linear regression. The presence of clinical
signs of osteoarthritis had a signicant effect on actigraphic recordings, with osteoarthritic dogs spending lower
proportions of the night period resting (and therefore higher proportions of the night period active) compared to
control dogs (z =2.21; P =0.0268). However, there was no signicant difference between the SNoRE scores of
osteoarthritic and control dogs (z = − 1.01, p =0.312). The actigraphic ndings of this study suggest that dogs
with osteoarthritis may experience impaired sleep, which could have important welfare implications and merits
further study.
1. Introduction
Osteoarthritis is a common cause of chronic pain in humans (Are-
ndt-Nielsen et al., 2010; Peat et al., 2001; Sofat et al., 2011), and human
osteoarthritis patients often report sleep impairments (Power et al.,
2005; Taylor-Gjevre et al., 2011; Wilcox et al., 2000). These disruptions
to sleeping patterns are likely to be associated with the chronic pain of
osteoarthritis, since chronic pain is a predictor of sleep problems in other
human conditions (Drewes et al., 2000; Nicassio and Wallston, 1992;
Riley et al., 2001). Sleep impairments themselves may exacerbate
chronic pain (Afeck et al., 1996; Morin et al., 1998; Stone et al., 1997)
and adversely affect cognition (Chee and Choo, 2004; Steenari et al.,
2003). They therefore represent a signicant threat to wellbeing.
Osteoarthritis also causes chronic pain in dogs (Brown et al., 2007,
2008; Conzemius et al., 2003; Hielm-Bj¨
orkman et al., 2009; Hunt et al.,
2018; Moreau et al., 2003; Wiseman et al., 2001), and canine osteoar-
thritis is thought to be highly prevalent (Anderson et al., 2018; Henrotin
et al., 2005; Johnston, 1997). However, it is not known whether oste-
oarthritic dogs also experience sleep decits. In a blinded
placebo-controlled study, Knazovicky et al. (2015) found that
meloxicam analgesia caused signicant improvements in scores on an
owner questionnaire designed to measure sleep quality of dogs; the
Sleep and Night-time Restlessness Evaluation (SNoRE) questionnaire.
However, there were no signicant differences in actigraphic measure-
ments of night-time activity between meloxicam-treated and baseline or
placebo-treated osteoarthritic dogs. This suggests that meloxicam
treatment improves owner-reports of how well their osteoarthritic dog is
sleeping, but does not alter night-time movement of the dogs. However
Knazovicky et al. (2015) did not include a control group of healthy dogs,
and therefore it is not known whether the observed effects of meloxicam
analgesia on reported sleep in osteoarthritic dogs were due to a reversal
of the effects of osteoarthritis on sleep or due to a non-specic effect of
the analgesic, and hence whether osteoarthritis is indeed associated with
impaired sleep in dogs.
The aim of this study was to explore whether dogs with clinical signs
of osteoarthritis also display impaired sleep compared to healthy control
dogs. Polysomnography can distinguish true sleep and wakefulness pe-
riods as well as different sleep stages and has been developed for use in
dogs by Kis et al. (2014), initially in a canine sleep laboratory with re-
searchers on hand to place electrodes, but more recently in dog owners’
* Corresponding author.
E-mail address: mike.mendl@bris.ac.uk (M. Mendl).
1
Present address: Highcroft Veterinary Referrals, 615 Wells Rd, Whitchurch, Bristol BS14 9BE, UK.
Contents lists available at ScienceDirect
Applied Animal Behaviour Science
journal homepage: www.elsevier.com/locate/applanim
https://doi.org/10.1016/j.applanim.2022.105661
Received 4 February 2022; Received in revised form 13 May 2022; Accepted 25 May 2022

Applied Animal Behaviour Science 253 (2022) 105661
2
homes allowing in situ night-time recordings of up to 3 h duration
(Reicher et al., 2021). Given the aim of recording sleep over a 28-day
period, it was considered unlikely that sufcient owner compliance
including regular and correct placement of electrodes could be achieved
each night across the study period in owners’ homes. Therefore, poly-
somnography was considered infeasible for this study and, instead, sleep
behaviour was assessed by measuring the proportion of the night-time
period spent resting using an actigraphic device; the FitBark™ activity
monitor system.
In humans, actigraphy generally has agreement rates with poly-
somnography of approximately 80%. However, its sensitivity is
considerably higher than its specicity, which can lead to poorer
agreement rates in populations where participants spend lower pro-
portions of the night period sleeping (Sadeh, 2011). Nevertheless, it has
accurately discriminated human participants with sleep impairments
from control participants in a range of studies (extensively reviewed by
Sadeh, 2011), indicating that actigraphic measurements are a useful
proxy measure of sleep in these populations. Although actigraphy in
companion dogs has not yet been systematically validated against pol-
ysomnography, its promise in human sleep studies and its ability to
remotely monitor dog activity in owners’ homes over several weeks,
made it the most appropriate technique for objective monitoring of dog
sleep in this study.
We used the FitBark actigraphy system (Knazovicky et al., 2015 used
Actical Activity monitors) as it has been specically designed for dogs, is
signicantly more affordable than most competitors, allowing the pur-
chase of one device per participating dog, is very easy for owners to use
and synchronise, and allows data to be uploaded remotely from the
device via Bluetooth. Consequently, it was unnecessary to perform
repeated visits to the owner’s home to download data. Whilst the FitBark
has not been fully validated in the peer-reviewed literature, it has been
used in previous studies to investigate the effects of a nutraceutical
product on activity levels (Di Cerbo et al., 2017) and the effect of dogs’
activity and rest patterns on those of their owners (Patel et al., 2017).
Alongside actigraphic measures of sleep quality, we assessed sleep
quality using the SNoRE questionnaire developed by Knazovicky et al.
(2015). We hypothesised that dogs with osteoarthritis would show
decreased sleep quality (increased SNoRE scores) relative to control
dogs, and would also show signicantly decreased proportions of the
night period spent resting.
2. Methods
Ethical approval for animal use and human (owner) participation
was obtained from the University of Bristol Animal Welfare and Ethical
Review Body (VIN/17/005) and University of Bristol Faculty of Health
Sciences Research Ethics Committee (Application Ref: 31623)
respectively.
2.1. Animals
Forty-one dogs were recruited via a social media campaign (using a
combination of the study’s Facebook Page https://www.facebook.co
m/dog.arthritis and posts on dog-related groups and pages based in
South West England) as well as poster and leaet placement within
veterinary clinics in Bristol and North Somerset. Inclusion criteria
required dogs to be between 5 and 11 years of age and weigh less than
12 kg, to be free from signs of age-related cognitive decline and pain
conditions other than osteoarthritis, and to have no signs or history of
health conditions that could cause study participation to impair their
health or welfare. Dogs receiving analgesic medication were not
excluded, the numbers of dogs receiving analgesia and the types and
frequency of analgesic treatment are shown in Table 1. Dogs were
assigned to groups (osteoarthritis (n =20) or control (n =21)) based on
clinical examination by a veterinary surgeon (MS) using a standardised
clinical checklist and a verbal history take from the owner about any
signs of osteoarthritis (stiffness, pain, slowing down during walks, dif-
culty jumping or climbing; see Table S1 in Supplementary Material).
Signalments of the dogs recruited are shown in Table 1. Participating
owners received a £ 10 gift card (John Lewis Partnership, London, UK)
following completion of the study. Sample size was based on that pre-
viously used in a similar study by Knazovicky et al. (2015) which
investigated the effects of NSAID-treated and placebo-treated osteoar-
thritic dogs on night-time accelerometry and SNoRE scores in a cross-
over design (N =19).
2.2. Apparatus
The FitBark system (FitBark Inc., Kansas City, MO) consisted of an
electronic activity monitor (electronic accelerometry device, 3.9 cm *
2.8 cm*1.2 cm) attached to each dog’s collar, a smartphone app
allowing owners to upload their dog’s data, and an online database from
which each dog’s data were downloaded.
2.3. Procedure
A clinical examination including an orthopaedic examination of all
appendicular joints was performed to assign the dogs to the osteoar-
thritic or healthy control group. A clinical history was also taken from
the owner to determine any signs of osteoarthritis or potential signs of
other health problems. Since a single researcher was responsible for
clinical examinations and data analysis, it was not possible to blind
analysis with respect to group. Body condition score (BCS) was
measured on a scale from 1 to 9 (WSAVA Nutritional Assessment
Guidelines Task Force Members, 2011).
Owners received an instruction sheet detailing how to use and charge
the FitBark activity monitor and smartphone app, as well as four iden-
tical sets of questionnaires. Each of these contained the SNoRE as well as
two previously-validated clinical questionnaires designed to assess the
Table 1
Signalments of recruited dogs in each group.
Variable Control Group Osteoarthritis Group
Sex
Female (neutered) 6 12
Male (neutered) 15 8
Breed Class:
Gundog 10 10
Crossbred 4 3
Hound 1 0
Pastoral 1 4
Terrier 2 1
Utility 1 0
Working 2 2
Analgesia provision
No 20 10
Yes 1 10
Analgesia frequency
None 20 10
Occasional 1 3
Daily 0 7
Analgesia type
None 20 10
Nonsteroidal analgesia only 1 7
Other analgesics provided
a
0 3
Season of data collection
Summer (May-August 2017) 12 13
Winter (November-February 2017) 9 7
Continuous variables
Age (years) 7.86 ±0.50 7.80 ±0.77
Body Condition Score 4.67 ±0.42 5.50 ±0.62
Data are expressed as counts for each categorical variable level and as means
with 95% condence intervals for each continuous variable.
a
Of the three dogs that received “other analgesics”, one received tramadol in
addition to nonsteroidal analgesia, one received gabapentin and paracetamol in
addition to nonsteroidal analgesia, and one received tramadol only.
M. Smith et al.

Applied Animal Behaviour Science 253 (2022) 105661
3
severity of pain in dogs; the Helsinki Chronic Pain Index (HCPI), which
has a single outcome score (chronic pain score (range 0–44):
Hielm-Bj¨
orkman et al., 2009) and the Canine Brief Pain Inventory
(CBPI), which has two validated outcome scores (CBPI Severity score
(range 0–10) and CBPI Interference score (range 0–10)) as well as a
single-question quality of life (CBPI QOL (range 0–4)) score (Brown
et al., 2007).
The CBPI and HCPI questionnaires were selected for this study over
alternative questionnaires relating to chronic pain in dogs for two main
reasons. Firstly, they were used by Knazovicky et al. (2015), allowing
ndings from this study to be easily compared with theirs. Furthermore,
both questionnaires have been extensively validated. Both were
formulated via discussion between veterinary professionals and dog
owners followed by removal of questions with low inter-item correla-
tions or no difference between osteoarthritic and control dogs, giving
good face and content validity (Brown et al., 2007; Hielm-Bj¨
orkman
et al., 2003). They also have good criterion validity, being correlated
with existing measures of lameness and quality of life (Brown et al.,
2007, 2009; Hielm-Bj¨
orkman et al., 2009), and construct validity, with
PCA being performed to identify constructs measured and signicant
differences in scores between healthy and osteoarthritic dogs (Brown
et al., 2007, 2009; Hielm-Bj¨
orkman et al., 2003). They also have high
internal consistency and repeatability scores (Brown et al., 2009; Hielm
Bj¨
orkman et al., 2009) and have shown signicant responsiveness to
carprofen analgesia (Brown et al., 2008; Hielm-Bj¨
orkman et al., 2009).
The main drawback of the HCPI is that the English translation has not
itself been validated and the translation from Finnish is slightly stilted,
which may affect owners’ responses. This is not an issue for the CBPI
which was originally written and validated in English.
Each set of questionnaires was marked with the day of the study that
they were to be completed (7, 14, 21 and 28), where the system was
initially set up on day 0. Owners also received a sleep diary sheet in
which they recorded the times at which they went to bed at night and got
up in the morning each day over the course of the 28-day study period.
Owners were able to record on the sheet any events that may have
affected the dog’s activity or rest on a particular day.
2.4. Data preparation
Dogs wore the activity monitors 24 h per day except when they
needed charging which was always done during daytime hours. Raw
activity results for each minute from the 28-day study period were
downloaded from the FitBark website’s online database for each dog.
These consisted of the date and time of each recording along with a
recorded activity value. Proprietary algorithms provided by FitBark
were used to determine whether the dog was in a state of “rest”, “ac-
tivity” or “play” (high-intensity activity) for each minute recorded.
Because this study focused predominantly on the distinction between
rest and activity, and because it was difcult to determine what “play”
recordings truly represented, “play” and “activity” were combined into a
single category such that for each minute a dog was either at “rest” or
“activity”.
The true night-time period was calculated for each dog as the period
between one hour after the owner’s reported mean bedtime and one
hour before the owner’s mean getting up time, following Knazovicky
et al. (2015). The range was 296–461 min. For each minute of data
recorded for each dog, the following values were entered: The day
(0–28) and week (1–4) of the recording, whether the recording was
made during the daytime or night-time period, and whether the dog was
resting or active (according to the algorithm provided). From this the
proportion of the night-time period that was spent resting was calculated
for each day of the study for each dog. Questionnaire scores (SNoRE,
HCPI and CBPI) were calculated for each week for each dog from owner
questionnaire responses. SNoRE and HCPI scores were calculated as the
sum of all individual question scores for each dog in each week. CBPI
Pain Severity Score was calculated as the mean of individual question
scores for CBPI questions 1–4, and CBPI Pain Interference Score was
calculated as the mean of individual question scores for CBPI questions
5–10 for each dog in each week.
2.5. Statistical analysis
Differences in age, body condition score and questionnaire scores
(HCPI Score, CBPI Severity Score, CBPI Interference Score and CBPI
Quality of Life (QOL) score) between groups were investigated using
Mann-Whitney U-tests. Holm-Bonferroni corrections (Holm, 1979) to
the p-value thresholds for signicance (
α
) were performed to account for
multiple testing.
Due to non-normality of residuals and a large number of outliers,
both outcomes (proportion of night-time period spent resting and SNoRE
score) were analysed via robust mixed effects linear models, using the
“rlmer()” function from the “robustlmm” package (Koller, 2016) in R (R
Core Team, 2014). This is a modied version of the “lmer()” function
from the package “lme4” (Bates et al., 2015), which increases robustness
to non-normality and the presence of outliers at the expense of decreased
asymptotic efciency (Koller, 2016) (a measure of model quality related
to the variance associated with parameter estimation) (Everitt and
Skrondal, 2002; pp. 24 and 149).
Univariable models are commonly performed to select factors for
inclusion in a nal model (for example: Alves et al., 2002; Bogaert et al.,
2005; Kooby et al., 2003). Because in this study, the interaction of each
factor with group was as important as the main effects, and adding only
the main effects to each initial model may cause the omission of a factor
with a signicant interaction with group from the nal model, each
initial model included not only the factor of interest but also group and
the interaction between each factor and group as xed effects (see
Tables S2 and S3 in Supplementary Material). “Dog” and “week” were
included as random effects. Initial models were performed for two
outcome variables; proportion of the night spent resting and SNoRE
score.
All xed effects where p <0.1 in initial models were selected for
addition to the nal model for each outcome variable, as using a stricter
threshold of p <0.05 often leads to omission of factors that would be
signicant in the nal model (Bursac et al., 2008). Where multiple
continuous variables were to be added to the nal model, Pearson cor-
relation tests were performed to assess whether these were signicantly
(p <0.05) correlated with each other prior to inclusion. If so, the var-
iable with the lowest p-value in the initial model was selected for
addition to the nal model, and the factors it signicantly correlated
with were omitted, in order to reduce the risk of collinearity (Dormann
et al., 2013).
Although the focus of this study was on sleep-related behaviour at
night, we also investigated whether control and osteoarthritic dogs
differed in the proportions of time that they spent resting during the day
(i.e. all times not in the night period), and whether this was correlated
with night-time resting behaviour.
Results are given as test statistics with degrees of freedom (where
appropriate), p-values, and
α
-values where these have been adjusted to
account for multiple testing. R values are included for correlational tests.
3. Results
3.1. Signalments and questionnaire scores
The signalments of recruited dogs are summarised in Table 1. Oste-
oarthritic dogs had signicantly higher HCPI, CBPI Severity and CBPI
Interference scores than control dogs, but no differences were detected
for age, body condition score or CBPI QOL score (Table 2; Fig. 1).
3.2. Proportion of night period spent resting
Results of initial screening models are shown in Table S2 (see
M. Smith et al.

Applied Animal Behaviour Science 253 (2022) 105661
4
Supplementary Material). Since only Group was signicant at the
p<0.1 threshold, the nal model contained only Group as a xed effect.
The results of the nal model are shown in Table 3. Dogs with chronic
pain from osteoarthritis spent signicantly (p <0.05) lower proportions
of the night period resting compared to healthy control dogs, as shown
in Fig. 2.
3.3. Relationship between night-time and daytime resting behaviour
Proportion of the daytime spent resting did not differ between con-
trol and osteoarthritic dogs (β= − 0.00457 ±0.00308, z =0.55,
p=0.550). A weak positive (Spearman) correlation was found between
average proportion of time spent resting during the day and night
(rho=0.391, p =0.04), suggesting that some dogs are more generally
active than others.
3.4. SNoRE score
Independent variables that had a signicant effect on SNoRE score in
initial models (p <0.1) were HCPI score, CBPI Severity score, CBPI
Interference score, CBPI QOL score and the interaction between group
and HCPI score, as shown in Table S3 (see Supplementary Material).
However, all of these variables were signicantly correlated, as shown in
Table S4 (see Supplementary Material). Therefore, HCPI score was
selected for addition to the model because it had the lowest p-value
associated with its effect on SNoRE score. The nal model therefore
contained HCPI score as well as group and the interaction between HCPI
score and group (as this was signicant at the p <0.1 threshold in the
initial model) as xed effects. The results of this model are shown in
Table 4. There was no signicant effect of group on SNoRE score overall,
however the effects of HCPI score and the interaction between HCPI
score by group interaction were statistically signicant. For both groups,
higher HCPI scores were associated with higher SNoRE scores (more
night-time restlessness and poorer sleep). However, the magnitude of
this relationship was greater for control dogs than for osteoarthritic
Table 2
Comparisons between osteoarthritis and control groups for clinical questionnaire outcome scores and signalment variables (age and body condition score), with Holm-
Bonferroni-adjusted
α
-values.
Variable Mann-Whitney U-Statistic p-value p-value rank Holm-Bonferroni
α
-value Signicant following Bonferroni-Holm correction?
CBPI Interference score 34 5.07 * 10
-6
1 0.00833 Yes
CBPI Severity score 42.5 1.12 * 10
-5
2 0.01 Yes
HCPI score 60 9.45 * 10
-5
3 0.0125 Yes
Body Condition Score 132 0.03255 4 0.0167 No
CBPI QOL score 247 0.04302 5 – No
Age 206.5 0.9359 6 – No
Fig. 1. Box plots showing the differences in questionnaire outcome scores be-
tween groups. Asterisks indicate a signicant difference (p <0.05) between
groups. Boxplots show medians with interquartile ranges. Whiskers represent
the highest and lowest values within 1.5 interquartile ranges of the upper or
lower quartile. Outliers beyond this range are represented with points.
Table 3
Results of nal model for which the outcome was proportion of the night period
spent resting.
95% condence limits
Variable Beta estimate Lower Upper z-value p-value
Group
Control Ref.
Osteoarthritis 0.9760 0.9540 0.9970 2.2140 0.0268
Ref. =reference category. P-values indicating signicance at
α
=0.1 are shown
in bold.
Fig. 2. The proportion of the night-time period spent resting for osteoarthritic
and healthy control dogs. Asterisks indicate a signicant difference (p <0.05)
between groups. Values are shown as means with 95% condence intervals.
Crosses indicate mean values for each individual dog (with horizontal jitter
applied for improved visualisation of individual points).
Table 4
Results of nal model for which the outcome was SNoRE score.
95% Condence
Limits
Main and interaction
effects
Beta
estimate
Lower Upper z-value p-
value
Control Ref.
Osteoarthritis 16.0958 0.0740 3503.1130 -1.0117 0.3117
HCPI score 1.1512 1.0717 1.2365 -3.8564 0.0001
HCPI score*Group
Interaction
0.9008 0.8169 0.9933 2.0944 0.0362
Ref. =reference category. P-values indicating signicance at
α
=0.1 are shown
in bold.
M. Smith et al.

Applied Animal Behaviour Science 253 (2022) 105661
5
dogs, as shown in Fig. 3.
4. Discussion
The signicantly higher HCPI, CBPI Severity and CBPI Interference
scores of osteoarthritic compared to control dogs suggest that osteoar-
thritic dogs were experiencing more pain and had more mobility
impairment than control dogs, as expected. This indicates that the
clinical examination used was successfully able to differentiate osteo-
arthritic and healthy control dogs.
According to actigraphy, osteoarthritic dogs spent a lower propor-
tion of the night period resting than control dogs, suggesting that they
spent less time asleep during the night. This is consistent with ndings
that human osteoarthritis patients report impaired sleep (Power et al.,
2005; Taylor-Gjevre et al., 2011; Wilcox et al., 2000), and the ndings of
Fielden et al. (2003) that human osteoarthritic patients displayed
signicantly improved actigraphic measures of sleep following total hip
replacement, indicating that osteoarthritis-related pain was the cause of
their poorer sleep prior to surgery. Whilst Leigh et al. (1988) found no
signicant differences in actigraphic measurements between human
osteoarthritic patients and healthy control volunteers, they did observe
a non-signicant trend for osteoarthritic participants to move around
more during the night than control participants (Leigh et al., 1988). In a
recent study of horses classied according to their lameness and inferred
orthopaedic disease status, there was also no apparent relationship be-
tween (automated recordings of) daily recumbency times and lameness
(Kelemen et al., 2021). On the other hand, a small study of eight
video-recorded hospitalised horses indicated that animals with more
severe osteoarthritis spent less time in lateral recumbency than those
with milder disease (Oliveira et al., 2022). Because this is the rst study
to compare actigraphic measures of sleep in dogs with osteoarthritis to
those without, our ndings should be considered exploratory. Further
research is warranted to conrm these effects are generalisable to the
wider population of companion animal dogs.
There was no signicant effect of group on SNoRE score, suggesting
that osteoarthritis may not impair dogs’ sleep quality as perceived by
their owners. It is also possible that the sample size in this study may
have been insufcient to detect an effect. However, a sample size
calculation based on mean (sd) SNoRE scores for osteoarthritic (20.13
(8.2)) and control dogs (18.14 (8.7)) indicated that future studies would
require a sample of c.280 dogs per group to detect a signicant differ-
ence (t-test, p <0.05 with 0.8 power) based on the small SNoRE score
effect size observed here. The signicant effect of HCPI score may
indicate that dogs that experienced more severe pain due to osteoar-
thritis had impaired sleep quality. However, this effect was not evenly
observed - HCPI score had a greater effect on SNoRE scores of control
dogs than osteoarthritic dogs. This could potentially be because the
subset of control dogs with elevated HCPI scores may have had another
undiagnosed pain-causing condition that affected their sleep quality
more than osteoarthritis did in the osteoarthritic dogs, potentially
masking an effect of osteoarthritis on sleep quality. Alternatively, it is
possible that owners of dogs with elevated scores in both the HCPI and
SNoRE questionnaires may have had a tendency to perceive their dogs’
behaviour more negatively than other owners, and thus tended to give
higher scores on both questionnaires.
Our ndings contrast with those of Knazovicky et al. (2015) that
meloxicam analgesia signicantly improved SNoRE scores, but not an
actigraphic measure of sleep (total night-time activity), in dogs with
osteoarthritis. There are several potential reasons for this. Different
devices and algorithms were used, which are known to cause differences
in actigraphic measures of sleep in humans (Paquet et al., 2007).
Recruitment criteria were different; in particular the present study had
no requirement for dogs to exhibit radiographic signs of osteoarthritis.
Since radiographic signs of osteoarthritis are not always closely associ-
ated with clinical severity (Gordon et al., 2003; Hielm-Bj¨
orkman et al.,
2003), it was considered that radiography would have been unneces-
sarily invasive, require owners to visit the veterinary hospital, and
would introduce unnecessary anaesthetic risks.
The present study also did not exclude dogs receiving analgesic
treatment, in order to avoid biasing recruitment such that dogs with
more severe pain from osteoarthritis that required analgesic treatment
would have been excluded from the study and dogs with less severe pain
from osteoarthritis that did not require analgesic treatment would not.
Other than analgesics, one female dog with osteoarthritis was receiving
Estriol for urinary incontinence and one control dog received pheno-
barbitone during the study period. The latter could potentially have
affected this dog’s behaviour but he had the seventh lowest mean pro-
portion of time spent resting during the night and the joint sixth lowest
mean SNoRE score in the control group (21 dogs), suggesting that his
night time resting behaviour was not markedly different from the other
control dogs.
The actigraphic outcome measures recorded were also different (the
present study used mean proportions of night-time spent resting whereas
Knazovicky et al., 2015 used mean activity counts per minute over the
night-time period), which may be affected differently by
osteoarthritis-related pain, since the night-time duration may have
varied between individual dogs. It is also possible that meloxicam
analgesia improves sleep in osteoarthritic dogs via alternative mecha-
nisms than simply reducing the effects of osteoarthritis on sleep, and
thus the differences in various measures of sleep between
placebo-treated osteoarthritic dogs and meloxicam-treated osteoar-
thritic dogs may not directly reect the differences between osteoar-
thritic and healthy control dogs. Additionally, intensity or duration of
activity during the preceding day may have affected subsequent
night-time activity, and we observed a weak positive relationship be-
tween the two suggesting that some dogs were more generally active
across a day.
It is likely that owners were not regularly present in the same room as
the dog whilst the dog was sleeping, and so may have guessed the an-
swers to the SNoRE or based these on the dog’s resting behaviour during
the day. This is also possible in the study performed by Knazovicky et al.
(2015) as dog or owner sleeping locations were not described in the
inclusion criteria. Furthermore, presence or absence of the owner in the
same room could have affected the dogs’ sleep. In our study we know
that all dogs slept indoors and, through informal communications dur-
ing data collection, that the majority slept in different rooms to their
owners, but we do not have quantitative data on the latter point. Whilst
there is some conicting evidence on how the presence of dogs can affect
human sleep (Brown et al., 2018; Patel et al., 2017; Smith et al., 2018), it
is currently unknown whether and how the presence of humans in-
uences canine sleep.
Fig. 3. The HCPI and SNoRE scores of dogs in each group. Points represent
mean observed values for each dog. Lines represent scores predicted by the
model, with shaded areas representing interquartile ranges for model estimates.
M. Smith et al.

Applied Animal Behaviour Science 253 (2022) 105661
6
Whilst it was possible to blind owners as to whether their dog
received placebo or nonsteroidal treatment in the study performed by
Knazovicky et al. (2015), it is less easy to thoroughly blind owners as to
whether their dogs had clinical signs of osteoarthritis, as most owners of
osteoarthritic dogs were already aware of osteoarthritis-induced
behavioural changes and described them within the dog’s clinical his-
tory. However, it would be expected that the effects of owner perception
would have caused an increased difference in SNoRE scores between
osteoarthritic and control dogs, rather than preventing such a difference
from being found. Additionally, because the SNoRE score has yet to be
thoroughly validated, it is possible that it was unable to detect differ-
ences in night-time restlessness between osteoarthritic and control dogs,
despite being able to detect differences between placebo-treated and
NSAID-treated osteoarthritic dogs in Knazovicky et al.’s (2015) study.
These caveats and the small treatment effect size on SNoRE score
observed here compared to our actigraphic ndings suggest that SNoRE
should be used with caution prior to further validation, especially when
owners have little information on their dog’s behaviour at night.
Furthermore, future studies should address a limitation of this study and
that of Knazovicky et al. (2015) by recording where dogs sleep and
whether owners are present, and controlling for these factors in statis-
tical analyses.
A further study limitation is that the researcher evaluating osteoar-
thritis (MS) also carried out the statistical analysis and was not blind to
treatment group identities. Questionnaire results were entered by
another researcher, but it was not possible for multiple veterinary staff
to go on home visits to collect osteoarthritis information independently
to MS. This would have been more feasible had the study been per-
formed on-site at Bristol Veterinary School, but this may have affected
dog behaviour during clinical examination and, from previous experi-
ence, would likely have impeded owner recruitment.
The actigraphic ndings from this study suggest that dogs with
osteoarthritis may experience impaired sleep, as seen in human patients
with osteoarthritis (Power et al., 2005; Taylor-Gjevre et al., 2011; Wil-
cox et al., 2000) and other chronic pain conditions (Nicassio and Wall-
ston, 1992; Riley et al., 2001). This could potentially be a welfare
concern because sleep disturbance is thought to exacerbate pain severity
in human chronic pain patients (Afeck et al., 1996; Morin et al., 1998),
therefore impaired sleep may also cause increased pain severity in
osteoarthritic dogs. It is not yet known whether the relatively small in-
creases in night-time activity observed in this study represent suf-
ciently impaired sleep to exacerbate dogs’ pain, but due to the potential
welfare impacts this is worthy of future investigation.
Impaired sleep is also associated with impaired working memory in
humans (Chee and Choo, 2004; Steenari et al., 2003), therefore sleep
impairments may affect the ability of dogs with osteoarthritis to solve
problems or navigate around their environment, which could alter their
ability to respond to training and to engage with the environment when
walking with owners. In line with these suggestions, recent dog studies
show that memory improvements in command learning tasks are related
to EEG spectral features and spindle density during pre-test sleep periods
(Kis et al., 2017; Iotchev et al., 2017, 2020a), and that EEG sleep spindle
frequency characteristics of individual dogs are associated with their
reversal learning ability (Iotchev et al., 2020b). These ndings suggest
that, as in humans, there are links between sleep and cognitive perfor-
mance in dogs and therefore that disrupted sleep may indeed have
detrimental effects on dog learning and memory.
Furthermore, since sleep disturbance in human chronic pain condi-
tions is often preceded by increased pain severity (Drewes et al., 2000;
Nicassio and Wallston, 1992; Riley et al., 2001), sleep disturbance in
canine osteoarthritis may also reect increased pain and hence be useful
as a welfare indicator. Because of these potential implications, further
research is needed to conrm whether the ndings of this study are
generalisable to the wider canine population.
5. Conclusions
Osteoarthritic dogs in this study spent less time resting during the
night than healthy control dogs, but did not have signicantly different
SNoRE scores. This suggests that dogs with osteoarthritis may spend less
time sleeping during the night than control dogs and may experience
impaired sleep similar to that reported in human osteoarthritis patients,
and with potentially negative implications for their welfare.
Conict of interest statement
The authors acknowledge no conicts of interest regarding this
study.
Data Availability
Data for this article are available on request from the authors.
Acknowledgements
This research was funded by the UK Biotechnology and Biological
Sciences Research Council (BBSRC) South West Biosciences Doctoral
Training Partnership (SWBio DTP) programme, grant number BB/
J014400/1. We are grateful to the owners of the dogs for their enthu-
siastic participation in the study, and to three anonymous referees for
their helpful and constructive comments.
Appendix A. Supplementary material
Supplementary data associated with this article can be found in the
online version at doi:10.1016/j.applanim.2022.105661.
References
Afeck, G., Urrows, S., Tennen, H., Higgins, P., Abeles, M., 1996. Sequential daily
relations of sleep, pain intensity, and attention to pain among women with
bromyalgia. Pain 68, 363–368.
Alves, A., Panis, Y., Trancart, D., Regimbeau, J.-M., Pocard, M., Valleur, P., 2002. Factors
associated with clinically signicant anastomotic leakage after large bowel
resection: multivariate analysis of 707 patients. World J. Surg. 26, 499–502.
Anderson, K.L., O’Neill, D.G., Brodbelt, D.C., Church, D.B., Meeson, R.L., Sargan, D.,
Summers, J.F., Zulch, H., Collins, L.M., 2018. Prevalence, duration and risk factors
for appendicular osteoarthritis in a UK dog population under primary veterinary
care. Sci. Rep. 8, 5641.
Arendt-Nielsen, L., Nie, H., Laursen, M.B., Laursen, B.S., Madeleine, P., Simonsen, O.H.,
Graven-Nielsen, T., 2010. Sensitization in patients with painful knee osteoarthritis.
Pain 149, 573–581.
Bates, D., M¨
achler, M., Bolker, B., Walker, S., 2015. Fitting linear mixed-effects models
using lme4. J. Stat. Softw. 1 (1).
Bogaert, D., Hermans, P.W.M., Boelens, H., Sluijter, M., Luijendijk, A., Rümke, H.C.,
Koppen, S., Belkum, Av, de Groot, R., Verbrugh, H.A., 2005. Epidemiology of
nasopharyngeal carriage of neisseria meningitidis in healthy dutch children. Clin.
Infect. Dis. 40, 899–902.
Brown, D.C., Boston, R.C., Coyne, J.C., Farrar, J.T., 2007. Development and
psychometric testing of an instrument designed to measure chronic pain in dogs with
osteoarthritis. Am. J. Vet. Res. 68, 631–637.
Brown, D.C., Boston, R.C., Coyne, J.C., Farrar, J.T., 2008. Ability of the canine brief pain
inventory to detect response to treatment in dogs with osteoarthritis. J. Am. Vet.
Med. Assoc. 233, 1278–1283.
Brown, D.C., Boston, R.C., Coyne, J.C., Farrar, J.T., 2009. A novel approach to the use of
animals in studies of pain: validation of the canine brief pain inventory in canine
bone cancer. Pain Med. 10, 133–142.
Brown, C.A., Wang, Y., Carr, E.C., 2018. Undercover dogs: pet dogs in the sleep
environment of patients with chronic pain. Soc. Sci. 7, 157.
Bursac, Z., Gauss, C.H., Williams, D.K., Hosmer, D.W., 2008. Purposeful selection of
variables in logistic regression. Source Code Biol. Med. 3, 17-17.
Di Cerbo, A., Sechi, S., Canello, S., Guidetti, G., Fiore, F., Cocco, R., 2017. Behavioral
disturbances: an innovative approach to monitor the modulatory effects of a
nutraceutical diet. J. Vis. Exp. 119, e54878.
Chee, M.W.L., Choo, W.C., 2004. Functional imaging of working memory after 24 hr of
total sleep deprivation. J. Neurosci. 24, 4560–4567.
Conzemius, M.G., Aper, R.L., Corti, L.B., 2003. Short-term outcome after total elbow
arthroplasty in dogs with severe, naturally occurring osteoarthritis. Vet. Surg. 32,
545–552.
Dormann, C.F., Elith, J., Bacher, S., Buchmann, C., Carl, G., Carr´
e, G., Marqu´
ez, J.R.G.,
Gruber, B., Lafourcade, B., Leit˜
ao, P.J., 2013. Collinearity: a review of methods to
M. Smith et al.

Applied Animal Behaviour Science 253 (2022) 105661
7
deal with it and a simulation study evaluating their performance. Ecography 36,
27–46.
Drewes, A.M., Nielsen, K.D., Hansen, B., Taagholt, S.J., Bjerregård, K., Svendsen, L.,
2000. A longitudinal study of clinical symptoms and sleep parameters in rheumatoid
arthritis. Rheumatology 39, 1287–1289.
Everitt, B., Skrondal, A., 2002. The Cambridge Dictionary of Statistics. Cambridge
University Press, Cambridge.
Fielden, J.M., Gander, P.H., Horne, J.G., Lewer, B.M.F., Green, R.M., Devane, P.A., 2003.
An assessment of sleep disturbance in patients before and after total hip arthroplasty.
J. Arthroplast. 18, 371–376.
Gordon, W.J., Conzemius, M.G., Riedesel, E., Besancon, M.F., Evans, R., Wilke, V.,
Ritter, M.J., 2003. The relationship between limb function and radiographic
osteoarthrosis in dogs with stie osteoarthrosis. Vet. Surg. 32, 451–454.
Henrotin, Y., Sanchez, C., Balligand, M., 2005. Pharmaceutical and nutraceutical
management of canine osteoarthritis: Present and future perspectives. Vet. J. 170,
113–123.
Hielm-Bj¨
orkman, A.K., Kuusela, E., Liman, A., Markkola, A., Saarto, E., Huttunen, P.,
Lepp¨
aluoto, J., Tulamo, R.-M., Raekallio, M., 2003. Evaluation of methods for
assessment of pain associated with chronic osteoarthritis in dogs. J. Am. Vet. Med.
Assoc. 222, 1552–1558.
Hielm-Bj¨
orkman, A.K., Rita, H., Tulamo, R.-M., 2009. Psychometric testing of the
Helsinki chronic pain index by completion of a questionnaire in Finnish by owners of
dogs with chronic signs of pain caused by osteoarthritis. Am. J. Vet. Res. 70,
727–734.
Holm, S., 1979. A simple sequentially rejective multiple test procedure. Scand. J. Stat.
65–70.
Hunt, J.R., Goff, M., Jenkins, H., Harris, J., Knowles, T.G., Lascelles, B.D.X.,
Enomoto, M., Mendl, M., Whay, B., Murrell, J.C., 2018. Electrophysiological
characterisation of central sensitisation in canine spontaneous osteoarthritis. Pain
159, 2318–2330.
Iotchev, I.B., Kis, A., B´
odizs, R., van Luijtelaar, G., Kubinyi, E., 2017. EEG transients in
the sigma range during non-REM sleep predict learning in dogs. Sci. Rep. 7, 12936.
Iotchev, I.B., Reicher, V., Kov´
acs, E., Kov´
acs, T., Kis, A., G´
acsi, M., Kubinyi, E., 2020b.
Averaging sleep spindle occurrence in dogs predicts learning performance better
than single measures. Sci. Rep. 10, 22461.
Iotchev, I.B., Szab´
o, D., Kis, A., Kubinyi, E., 2020a. Possible association between spindle
frequency and reversal-learning in aged dogs. Sci. Rep. 10, 6505.
Johnston, S.A., 1997. Osteoarthritis: joint anatomy, physiology, and pathobiology. Vet.
Clin. N. Am. Small Anim. Pract. 27, 699–723.
Kelemen, Z., Grimm, H., Long, M., Auer, U., Jenner, F., 2021. Recumbency as an equine
welfare indicator in geriatric horses and horses with chronic orthopaedic disease.
Animals 11, 3189.
Kis, A., Szakad´
at, S., G´
acsi, M., Kov´
acs, E., Simor, P., T¨
or¨
ok, C., Gombos, F., B´
odizs, R.,
Top´
al, J., 2017. The interrelated effect of sleep and learning in dogs (Canis
familiaris); an EEG and behavioural study. Sci. Rep. 7, 41873.
Kis, A., Szakad´
at, S., Kov´
acs, E., G´
acsi, M., Simor, P., Gombos, F., Top´
al, J., Mikl´
osi, ´
A.,
B´
odizs, R., 2014. Development of a non-invasive polysomnography technique for
dogs (Canis familiaris). Physiol. Behav. 130, 149–156.
Knazovicky, D., Tomas, A., Motsinger-Reif, A., Lascelles, B.D.X., 2015. Initial evaluation
of nighttime restlessness in a naturally occurring canine model of osteoarthritis pain.
PeerJ 3, e772.
Koller, M., 2016. robustlmm: an R package for robust estimation of linear mixed-effects
models. J. Stat. Softw. 75, 1–24.
Kooby, D.A., Stockman, J., Ben-Porat, L., Gonen, M., Jarnagin, W.R., Dematteo, R.P.,
Tuorto, S., Wuest, D., Blumgart, L.H., Fong, Y., 2003. Inuence of transfusions on
perioperative and long-term outcome in patients following hepatic resection for
colorectal metastases. Ann. Surg. 237, 860.
Leigh, T.J., Hindmarch, I., Bird, H.A., Wrisght, V., 1988. Comparison of sleep in
osteoarthritic patients and age and sex matched healthy controls. Ann. Rheum. Dis.
47, 40.
Moreau, M., Dupuis, J., Bonneau, N.H., Desnoyers, M., 2003. Clinical evaluation of
nutraceutical, carprofen and meloxicam for the treatment. Vet. Rec. 152, 323–329.
Morin, C.M., Gibson, D., Wade, J., 1998. Self-reported sleep and mood disturbance in
chronic pain patients. Clin. J. Pain 14, 311–314.
Nicassio, P.M., Wallston, K.A., 1992. Longitudinal relationships among pain, sleep
problems, and depression in rheumatoid arthritis. J. Abnorm. Psychol. 101, 514.
Oliveira, T., Santos, A., Silva, J., Trindade, P., Yamada, A., Jaramillo, F., Silva, L.,
Baccarin, R., 2022. Hospitalisation and disease severity alter the resting pattern of
horses. J. Equine Vet. Sci. 110, 103832.
Paquet, J., Kawinska, A., Carrier, J., 2007. Wake detection capacity of actigraphy during
sleep. Sleep 30, 1362–1369.
Patel, S.I., Miller, B.W., Kosiorek, H.E., Parish, J.M., Lyng, P.J., Krahn, L.E., 2017. The
effect of dogs on human sleep in the home sleep environment. Mayo Clin. Proc. 92,
1368–1372.
Peat, G., McCarney, R., Croft, P., 2001. Knee pain and osteoarthritis in older adults: a
review of community burden and current use of primary health care. Ann. Rheum.
Dis. 60, 91–97.
Power, J.D., Perruccio, A.V., Badley, E.M., 2005. Pain as a mediator of sleep problems in
arthritis and other chronic conditions. Arthritis Care Res. 53, 911–919.
R Core Team, 2014. R: A Language and Environment for Statistical Computing. R
Foundation for Statistical Computing, Vienna, Austria.
Reicher, V., Bunford, N., Kis, A., Carreiro, C., Csibra, B., Kratz, L., Gacsi, M., 2021.
Developmental features of sleep electrophysiology in family dogs. Sci. Rep. 11,
22760.
Riley, J.L., Benson, M.B., Gremillion, H.A., Myers, C.D., Robinson, M.E., Smith Jr, C.L.,
Waxenberg, L.B., 2001. Sleep disturbance in orofacial pain patients: pain-related or
emotional distress? CRANIO® 19, 106–113.
Sadeh, A., 2011. The role and validity of actigraphy in sleep medicine: an update. Sleep
Med. Rev. 15, 259–267.
Smith, B.P., Browne, M., Mack, J., Kontou, T.G., 2018. An exploratory study of
human–dog co-sleeping using actigraphy: do dogs disrupt their owner’s sleep?
Anthrozo¨
os 31, 727–740.
Sofat, N., Ejindu, V., Kiely, P., 2011. What makes osteoarthritis painful? The evidence for
local and central pain processing. Rheumatology 50, 2157–2165.
Steenari, M.-R., Vuontela, V., Paavonen, E.J., Carlson, S., Fj¨
allberg, M., Aronen, E.T.,
2003. Working memory and sleep in 6-to 13-year-old schoolchildren. J. Am. Acad.
Child Adolesc. Psychiatry 42, 85–92.
Stone, A.A., Broderick, J.E., Porter, L.S., Kaell, A.T., 1997. The experience of rheumatoid
arthritis pain and fatigue: examining momentary reports and correlates over one
week. Arthritis Rheumatol. 10, 185–193.
Taylor-Gjevre, R.M., Gjevre, J.A., Nair, B., Skomro, R., Lim, H.J., 2011. Components of
sleep quality and sleep fragmentation in rheumatoid arthritis and osteoarthritis.
Musculoskelet. Care 9, 152–159.
Wilcox, S., Brenes, G.A., Levine, D., Sevick, M.A., Shumaker, S.A., Craven, T., 2000.
Factors related to sleep disturbance in older adults experiencing knee pain or knee
pain with radiographic evidence of knee osteoarthritis. J. Am. Geriatr. Soc. 48,
1241–1251.
Wiseman, M.L., Nolan, A.M., Reid, J., Scott, E.M., 2001. Preliminary study on owner-
reported behaviour changes associated with chronic pain in dogs. Vet. Rec. 149,
423–424.
WSAVA Nutritional Assessment Guidelines Task Force Members, 2011. WSAVA
nutritional assessment guidelines. J. Feline Med. Surg. 13, 516–525.
M. Smith et al.