Physical activity (PA) is a well-recognised indicator and determinant of health [1] and more recently, sedentary behaviour (SB, sitting or lying with low energy expenditure whilst awake; [2]) has been identified as an independent risk factor for poor health [3]. Nonetheless, it is clear that whilst overall PA level decreases with age [4], older adults (>65 years) are also normally the most sedentary section of the population [5]. For adults, including older adults, national PA guidelines recommend 150 min per week of moderate to vigorous physical activity (MVPA; PA which raises the heart rate) [6]. Although a reduction in time spent in prolonged sitting is also recommended, there is no consensus about optimal sitting times. Maintaining appropriate levels of PA, or working towards this target at a later stage in life, has been shown to have significant health benefits for people with and without disease, and forms the focus for interventions targeting older adults. Higher levels of PA, including walking, are associated with improved health [7], reduced mortality, independent living, maintaining effective function and improved quality of life [8].

Dog ownership, and in particular dog walking as a feature of ownership, has been shown to be related to overall PA levels in a range of age groups. A meta-analysis of 29 studies conducted over 20 years examining the activities of Dog Owners (DO) and Non-Dog Owners (NDO) for a wide range of participants, including older adults, concluded that DOs walked more and were more physically active than NDOs [9], mostly from self-report PA measures. For example, post-menopausal female DOs were more likely to self-report 150 min per week exercise and less likely to be sedentary [10]. Older adult DOs (n = 330) engaged in more self-reported walking, not specifically intended as exercise (68 min/week), than non-pet owners (32 min/week) or non-dog pet owners (32 min/week) [11], but there was no significant difference between the groups in time spent walking for exercise (75, 62 and 52 min/week, respectively). The use of self-report means the robustness of these apparent effects in older adults can be questioned due to issues such as recall bias and social desirability bias. This can be a particular problem for studies investigating the effect of dog walking, as walking the dog may be a regular, planned activity that is easier to recall than other incidental PA, and also seen as something that owners should do, for example as a moral duty or to ensure the welfare of the animal. Objective measures of PA and SB provide opportunities to gain insight into both the intensity and pattern of PA and SB, allowing closer scrutiny of the potential relationship between dog ownership and health. Adult DOs who walked their dogs [12] had significantly longer time in MVPA (ActiGraph (ActiGraph Corp, Pensacola, FL, USA), 35 ± 24 min/day) and were more likely to meet PA recommendations (53%) than either NDOs (33 ± 24 min/day; 46%) or DOs who did not walk their dogs (27 ± 21 min/day; 33%). Older adults who walked their dog took approximately 1700 more steps (ActiGraph) than those who did not walk a dog [13].

A variety of monitors exist for objective measurement of physical activity and sedentary behaviour, with different strengths and weaknesses. The Actigraph monitor is worn at the hip and uses a threshold of low movement to identify SB, meaning that some activities undertaken while standing, such as washing the dishes, can be misclassified as sedentary tasks [14]. By contrast, the activPAL (PAL Technologies Ltd., Glasgow, UK) is worn on the front of the thigh and uses the static component of gravity to distinguish sitting and lying postures from standing, and is generally held as the gold standard for measuring SB [15]. The step count function of the Actigraph can underestimate total steps by as much as 40% at normal walking speeds (0.89 m/s [16]), whereas step count measured by the activPAL in older adults has been reported to be >99% accurate at similar speeds (≥0.67 m/s [17]). The activPAL is worn continuously, including overnight and during bathing (since it can be waterproofed), so that all PA and SB is measured, unlike the Actigraph, which is typically removed overnight and possibly at other times, increasing the likelihood that activity is missed.

Therefore, the aim of the current study was to use the activPAL monitor within a longitudinal design, in order to evaluate the association of dog ownership with both PA and SB in independently-mobile, community-dwelling older adults. Due to the potential for complex relationships between physical behaviour, dog ownership status and health, we chose to explore a range of health related physical activity and sedentary behaviour outcome measures. We hypothesised that owning a dog would be associated with increased physical activity (longer time spent walking, more steps taken) and reduced sedentary behaviour (less time spent sitting, less prolonged sitting, more sit-to-stand transitions).

Progress of the participants through the study is reported in Fig. 1. In total, 19 participants withdrew from the study, but in two of the pairs it was possible to replace the withdrawn participant with a different participant who was recruited and awaiting a match and who also met the matching criteria. The number of participants with data included in analysis was relatively similar across all sampling intervals (March–June, n = 77; July–October, n = 59; November–February, n = 82), with slightly lower levels of data collection over the summer.

Among those included in data analysis, compliance with wearing the monitor was excellent, with 92% (201/218 included datasets) having the full seven days of data. In total just 2% of potential days of data were not available for analysis. Most of the wake times used in the analysis (1310/1499 days, 87%) were derived from estimated wake and sleep times reported in the diary, with a further 11% (166 days) derived from diary reported bed times. Just 2% (34 days) of wake times were derived from the activPAL record.

Information on dogs owned was missing for two DOs. The remaining 43 DOs in this study owned 61 dogs, in total. Most DOs (n = 41, 95%) owned either one (n = 31, 72%) or two (n = 10, 23%) dogs, while the other two DOs owned four and six dogs. About half of the DOs reported that they were solely responsible for care (n = 17, 40%) and exercise (n = 23, 53%) of their dog(s). Of those DOs who shared responsibility for their dog(s), only a small number reported providing less than half of that care (n = 2) and exercise (n = 1). About half the dogs were female (n = 31, 49%), and most (n = 46, 75%) were neutered. Dogs were aged 7.7 ± 3.7 (0.3–15.0) years, and had been owned for most of their life 6.0 ± 3.7 (0.2–15.0) years. Most of the dogs owned were pedigree (n = 42, 69%), and the dogs were spread across a range of sizes (n = 20, 33% toy and small; n = 23, 38% medium; n = 18, 30% large and giant). Just under half of the dogs (n = 29, 46%) were usually walked on the lead.

DOs walked for significantly longer than NDOs overall and at a moderate cadence (Table 2). Consequently, DOs took significantly more steps than NDOs. The difference between the groups in time spent walking at least at a moderate cadence, of 21 min with 95%CI (12, 34 min), was similar to the difference in total time walking, 23 min with 95%CI (12,36 min), suggesting that the additional walking performed by the DOs was at a moderate cadence. Across all three data collection periods, significantly more DOs (87% (95%CI 61, 96)) than NDOs (47% (95%CI 19, 77)) met physical activity guidelines of 150 min of moderate activity per week (OR 75 (95%CI 3, 2167), p = 0.015). There was no significant difference between groups in time spent standing (Table 2). DOs and had fewer sedentary events, however there were no significant differences between the groups for time sedentary in total, the number or the duration of prolonged sedentary events (Table 2).

In this study owning a dog indicated a large, potentially health improving effect [25]: on average 20 min of additional time spent walking and 2700 additional steps per day with this additional walking undertaken at a moderate cadence (≥100 step/min). Indeed, the size of the difference may be sufficient to meet PA guidelines on its own (22 min of MVPA every day would achieve 150 min of MVPA per week). It is not surprising, therefore, that DOs were more likely to meet PA guidelines (87%, 95%CI 61, 96) than NDOs (47%, 95%CI 19, 77). Additionally, older adult DOs had 8 fewer sedentary events on average, but there was no difference in time spent sitting in total, in prolonged sedentary behaviour, or in time spent standing between the groups.

Previously reported group differences in MVPA in adults and adolescents, based on use of the ActiGraph were statistically significant, but often small (~2 min/day; [12, 26]). A small difference of 2 min/day is unlikely to have a large impact on health. The current study also found a larger difference in total step count (2700 steps) associated with dog ownership, than the difference found in the only other comparable study (1700 [13]). Although differences in total step count on their own do not provide information about the intensity at which they were taken, the PA guidelines could be achieved by taking 2200 steps per day (22 min per day, at a moderate cadence of 100 steps/min). Both studies, therefore, indicate a meaningful increase in walking due to dog ownership. In general, the NDOs in this group of older adults (7200 steps/day; 96 min/day walking) were within, but towards the top of, normal ranges, while the DOs (10,000 steps/day, 119 min/day walking) often exceeded usual ranges of PA compared to the general population of community dwelling older adults, (30–60% meeting guidelines, measured using self-report and ActiGraph [27, 28]).

Although not-significant, this study demonstrated a reduction in objectively measured SB of 19 min associated with dog ownership. Although there are no firm guidelines on SB and health, a dose relationship is apparent [3]. For example, a 1 h reduction in self-reported SB was equivalent to a 3% reduction in mortality in older women [29]. The scale of the difference associated with dog ownership in this study (19 min less time spent sedentary per day), is therefore likely to have a small positive influence on health. The only other study to assess objectively measured SB, used the ActiGraph and found a smaller and non-significant reduction in SB of 7 min/day in adolescents associated with household dog ownership [26]. Differences in measured sedentary behaviour may be due to different monitors used (assessing low movement versus posture), measurement protocol (removed overnight versus reported sleep), or differences in population behaviour (adolescents versus older adults). Of particular interest, is the composition of activity over a 24-h day. Given that the duration of a day is constant, a change in time spent in one type of activity, must result in a consequent and opposite change in time spent in other activities. In this study, there was little difference between the groups in time spent standing or time spent asleep, which implies that the reduction in time spent sedentary for the DOs (19 min reduction) was transferred into time spent in MVPA (21 min increase). Although the inter-relationship of sleep, SB and PA across the day is complex and under-researched [30], recent isotemporal time substitution analysis indicates that transfer of time spent sedentary to time spent in MVPA provides the maximum potential benefit to health [31].

Dog ownership represents a complex behavioural relationship between the DO, the dog(s), and other members of the household (including other pets). Owning a dog does not necessarily mean that an individual either cares for, or walks with, the dog. In some studies, dog walking, as opposed to dog ownership is assessed (e.g. [13]), and the general consensus in the literature is that it is dog walking, rather than dog ownership, which positively influences PA. In this study, the factor distinguishing groups was dog ownership, but we also assessed levels of self-reported caring for the dog. Most of the DOs in this study reported having sole responsibility for care of and walking the dog, with only one DO reporting less than 50% responsibility for walking the dog. A further factor that may influence the relationship of dog ownership and dog walking is the number and type of dog(s) owned. Factors which may influence this include gender, age, size [32, 33], breed, neuter status, temperament, energy [32] and behaviour of the dog [33, 34]. Theoretically, walking dogs on the lead may involve the DO walking slowly and stopping frequently, or dogs may get additional exercise when not on the lead. These factors should however be explored in more depth in future.

Although this study provides the best quality data to date on the effects of dog ownership on PA, there are several potential limitations to consider. Participants were volunteers and so may have been more physically active than the general population, so the results may not be completely generalisable. Any potential bias in the volunteers would apply to the NDOs as well as the DOs, and both groups had similar levels of health for the aspects we assessed, and so the results are valid for indicating the magnitude of effect. It should also be acknowledged that few volunteers were recruited from the lowest quintile of SES, and all volunteers had white British ethnicity, which may also limit the applicability of the findings to wider contexts. We did not assess the effect of confounders within our statistical models, however the matching between pairs of participants was excellent for common confounders, and therefore this is unlikely to have had a large influence on the difference between groups. The study was only powered to be able to detect changes in physical activity, and it is possible that a larger study would have detected a significant difference in sedentary behaviour outcomes. The design of this study does not allow any inference to be made about whether more active people are likely to own dogs, or whether DOs become more active through owning a dog.

A considerable strength of the study was use of an appropriate objective monitor to measure PA and SB. The activPAL monitor is considered a gold standard for measuring SB [15], and is accurate for measuring step count in older adults at normal walking speeds [17]. The monitor is also able to assess MVPA, using a threshold of cadence generally held to be at a moderate level. Because this is applied to the average cadence across an entire walking event, regardless of length, this assessment avoids issues involved with dividing PA data into arbitrarily defined units [35]. Although the ActiGraph monitor is generally held to be a good measure of MVPA, thresholds derived from laboratory based calibration studies are required to derive time spent in MVPA from hip acceleration [36]. A variety of different thresholds exist [30], differing between adults [e.g. 12] and children [e.g. 26], which can limit the ability for comparison between studies. Compliance with monitor wear and study protocols was high, leading to 92% of participants having a full 7 days data, this probably reflects our individualized recruitment strategy, and the ease of monitor wear, compared to larger generic studies. This compares favourably with other studies, for example, in the 2003–2004 National Health and Nutritional Examination Survey, only 67% of returned monitors had at least 4 days of data available for analysis [4]. As the monitor was worn continuously, including overnight and during water-based activities we were able to assess all PA and SB undertaken throughout the week. Measuring participants three times throughout a year allowed a robust assessment of whether a difference in PA is apparent throughout the year. Previous cross-sectional research into the influence of dog ownership on PA has generally not matched participants, instead breaking a group of recruited individuals into sub-groups based on their dog ownership status. In this study pairs of DOs and NDOs were matched on a range of basic characteristics (age, gender, ethnicity, SES, and cat ownership). Whilst there is a wide range of influences on both PA and SB behaviour, which may act on different levels (e.g. individual, environment, policy) [37], matching groups for some of the more basic characteristics has allowed the influence of dog ownership on PA and SB to be more effectively isolated.

This study found that older adult DOs walked on average 20 min a day longer than NDOs. These results confirm previous studies where DOs reported more walking than NDOs, but also indicate that the additional walking of DOs was undertaken at a moderate cadence. On average, DOs met recommended public health guidelines (30 min/day of moderate PA), but NDOs did not. Owning a dog, may therefore motivate older adults to engage in appropriate levels of PA for health. Health promotion professionals could consider encouraging appropriate dog ownership, or shared care of a dog to promote PA in older adults. The scale of the influence of dog ownership on PA found in this study, indicates that future research regarding PA in older adults should assess and report dog ownership and/or dog walking status. It is important to note that even if dog ownership is not the focus of a piece of research examining PA in older adults it may represent an important explanatory factor which should not be ignored.