Associations Between Gait Speed and Cognitive Domains in Older People with Cognitive Impairment

Introduction

Gait speed is an early predictor of cognitive impairment and dementia [1] and deteriorates with progression and severity of impairment [2]. Dementia disorders such as Alzheimer’s disease and vascular dementia result in significant disability and dependence in older people, and challenge health care providers worldwide [3]. A greater understanding of the relationship between gait and cognitive function may assist in optimizing interventions in this population.

While gait speed and stability have been linked with several cognitive domains [4], executive function may be the most important for gait control [5, 6] as it comprises higher-level cognitive processes that regulate behavior and responses essential for complex motor tasks. However, the few studies that have contrasted the strengths of relationships between gait speed and particular cognitive domains in people with dementia have reported conflicting results [7–10]. This may be because these studies have been small, and have not simultaneously adjusted for cognitive domains or adequately controlled for comorbidities. Further studies examining the relative importance of cognitive domains for gait control in larger samples of people with dementia are therefore warranted.

The aim of this study was to investigate the association between gait speed and cognitive function, including cognitive domains, in a large sample of community-dwelling older people with cognitive impairment. We hypothesized that gait would be associated with global cognition as well as several cognitive domains, but that executive function would have the strongest association.

Materials and methods

This cross-sectional study examined baseline data of 309 community-dwelling participants with cognitive impairment from the Intervention – Falls in Older Cognitively Impaired Subjects (iFOCIS) study [11]. The iFOCIS study, a fall prevention randomized controlled trial comprising tailored exercise and home hazard reduction programs, was conducted in Sydney, Australia. The study recruited participants aged 65 years or older from a range of clinical settings. All participants had a specialist clinical diagnosis of cognitive impairment, and/or scored <24 points on the Mini-Mental State Examination (MMSE) or the Mini-Addenbrooke’s Cognitive Examination, Australian Version, or <83 points on the Addenbrooke’s Cognitive Examination-III, Australian Version (ACE-III) [12]. Furthermore, all participants had a carer who had face-to-face contact for at least 3.5 hours per week. Exclusion criteria included MMSE <12, an inability to walk more than one meter despite assistance with a walking aid and/or another person, resident of care facility, blindness, severe psychiatric disease, a progressive neurological disease other than dementia or any medical condition precluding exercise (e.g., unstable cardiac disease). Written informed consent was obtained from either the participant or person responsible depending on a capacity assessment performed at the time of consent [11].

Ethical approval was obtained from the South Eastern Sydney Local Health District (HREC 14/046), human research ethics committee.

All assessments took place in participants’ homes by trained assessors. Demographic data, including past medical history and medication use, were self-reported with the assistance of the person responsible/caregiver. Gait speed in meters/second (m/s) was assessed in a single trial over a test distance of 2.4 meters. The distance was measured using a tape measure, with approximately one meter at each end to allow for acceleration and deceleration. Participants were instructed to walk at a comfortable pace with or without a walking aid. Gait speed over short distances has been shown to have good test-retest reliability in people with mild to moderate dementia (intraclass correlation coefficient = 0.86) [13]. Global cognitive function was assessed using the ACE-III, and has excellent reliability in people with dementia (Cronbach’s α coefficient = 0.88) [14]. The ACE-III comprises six domains: attention (0–18), memory (0–26), verbal fluency (0–14), language (0–26), and visuospatial ability (0–16); with a total score of 100 indicating highest function. Phonemic/letter and categorical/animal fluency, which involves enumerating as many words as possible in 60 s (words/min), test executive function and both contribute to the overall ACE-III verbal fluency score. In addition, the Frontal Assessment Battery (FAB) was completed to further assess executive function [15]. The FAB comprises six subtests that explore cognitive control processes responsible for appropriate goal-directed behavior and response to challenging situations. Each subtest scores from 0–3, with a total maximum score of 18 indicating highest function. The FAB has good to excellent reliability (Cronbach’s α coefficient = 0.78 and Cohen’s kappa coefficient = 0.87) in people with dementia [16]. Depressive symptoms were measured using the 15-item Geriatric Depression Scale (0–15) [17]. Physical activity was assessed using the Incidental and Planned Exercise Questionnaire (IPEQ) and expressed as hours of activity per week [18].

Results

The participants (n = 309) had a mean age of 82 years and 144 (47%) were female. Participants walked with a (mean±standard deviation, SD) gait speed of 0.62±SD 0.23 m/s, and 77 (25%) used a walking aid when gait speed was assessed (Table 1). Participants had a mean ACE-III score of 62.8±19.3 and a mean FAB score of 11.4±4.6 (Table 1).

Results from separate multiple linear regression analyses of associations between global and individual domains of cognitive function and gait speed are presented in Table 2. In minimally adjusted models (adjusted for age, sex, and education), global cognitive function (ACE-III), attention, verbal fluency, language, visuospatial ability, and executive function (FAB) were each associated with gait speed (Table 2). Executive function, assessed by verbal fluency and FAB, had the strongest association and explained the greatest variance in gait speed (adjusted R² = 0.188 and 0.229, respectively; Table 2). When the analyses were comprehensively adjusted (additionally adjusted for falls in the previous year, heart disease, hypertension, arthritis, depression, number of medications, walking aid use, and the Incidental and Planned Exercise Questionnaire) global cognitive function and each cognitive domain were significantly associated with gait speed (Table 2). Executive function (verbal fluency and FAB) and visuospatial ability demonstrated the strongest association and explained the greatest variance in gait speed (adjusted R²  = 0.401, 0.431, and 0.404. respectively). Overall, memory had the weakest association and the FAB the strongest with gait speed in both minimally and comprehensively adjusted analyses.

Results from analyses of associations between executive function and gait speed, adjusted separately for each individual cognitive function domain are presented in Table 3. Verbal fluency remained significantly associated with gait speed when adjusted for attention, memory, language, and visuospatial ability (Table 3). In contrast, visuospatial ability was the only cognitive domain to withstand adjustment for verbal fluency (Table 3). Similarly, the association between the FAB and gait speed remained statistically significant when adjusted for each ACE-III cognitive domain (excluding verbal fluency; Table 3). Conversely, none of the ACE-III cognitive domains withstood adjustment for the FAB (Table 3).

Discussion

This study examined the relationship between cognitive performance, specifically global cognition and individual cognitive domains, and gait speed in a large cohort of older people with cognitive impairment. We found global cognition and each cognitive domain were associated with gait speed, but that executive function, measured by the FAB, explained the greatest variance in gait speed.

In healthy older people, gait speed is associated with global cognition [4, 7]. Our findings reveal a similar relationship in people with cognitive impairment that is independent of possible confounding factors including age, sex, education, history of falls, heart disease, hypertension, arthritis, depression, number of medications, walking aid use, and level of physical activity. The observed association between gait speed and global cognition is in agreement with two previous smaller studies of people with early dementia, where associations were compared in adjusted analysis limited to age, sex, and education [9, 10]. In contrast, another cross-sectional study (n = 161) involving people with dementia living in nursing homes, comprehensively adjusted analysis did not support an association between global cognition and gait speed [8]. This latter finding could in part be attributed to the higher prevalence of physical impairment of nursing home residents, as indicated by the larger proportion of individuals using walking aids.

Consistent with previous studies of people with cognitive impairment [19] and dementia [8, 20], we found an association between gait speed and executive function. In our study, executive function (as measured with the FAB) demonstrated the strongest association with gait speed, as indicated by the value of the standardized regression co-efficient and adjusted R² in both minimally and comprehensively adjusted models. An association between attention and gait speed has also been reported in people with mild cognitive impairment [21], and was confirmed in our study. In addition, we found an association between memory, language, and visuospatial domains and gait. Such associations have not previously been demonstrated in people with cognitive impairment or dementia, but have been observed in general older populations [4, 7].

The strength of the relationship between executive function and gait speed was reinforced by the additional analysis of ACE-III domains, where executive function remained associated with gait in each model, independent of adjustment for other domains. In contrast, in the same analysis the association between gait speed and attention, memory, and language were attenuated. A similar result has been demonstrated in a study of general older adults, where gait pace (as opposed to rhythm) was shown to be independently associated with executive function in multiple analysis adjusted for cognitive domains including memory and processing speed [6]. The association between gait speed and visuospatial ability also withstood adjustment for verbal fluency but not FAB. This may indicate that visuospatial processing is important for gait control, but may also indirectly reflect the importance of executive function, as the visuospatial tasks included copying and drawing images (constructional praxis) [12].

The mechanisms linking gait with cognition remain elusive, but shared neural pathways vulnerable to normal aging as well as the pathologies associated with dementia have been suggested [22, 23]. The shared circuitry hypothesis means physical exercise may reduce or delay the decline in cognitive performance and/or incidence of dementia [24]. Limited evidence supports physical exercise as a means to slow cognitive decline in individuals with established cognitive impairment or dementia, but more robust, high quality trials are needed [25, 26]. Combined interventions targeting both gait and cognition may also be a potential treatment in individuals with cognitive impairment/dementia, and recent evidence has demonstrated positive cognitive outcomes [27]. In addition, pilot studies have shown that cognitive exercise targeting executive function can improve gait in older adults [28, 29]. Given the association between executive function and gait speed, longitudinal studies exploring their temporal relationship are warranted. Such studies could potentially identify strategies to maintain either or both executive function and gait in older age.

This study has both strengths and limitations. While the cross-sectional design of the study precludes causal inferences, analyses were comprehensively adjusted for potential confounders to reduce bias and increase internal validity. The inclusion criteria prevented analysis based on type of dementia, for example Alzheimer’s disease. Although collinearity assumptions were met, there may be some degree of overlap between cognitive domains, particularly in relation to executive function as this domain is difficult to measure independently of other cognitive domains [30]. For example, in addition to executive function, verbal fluency may also measure verbal ability, semantic memory and processing speed. Furthermore, as indicated above, the visuospatial domain involves constructional praxis, which also requires executive function [12].