Abstract
Distinguishing dementia subtypes can be difficult due to similarities in clinical presentation. There is increasing interest in discrete gait characteristics as markers to aid diagnostic algorithms in dementia. This structured review explores the differences in quantitative gait characteristics between dementia and healthy controls, and between four dementia subtypes under single-task conditions: Alzheimer’s disease (AD), dementia with Lewy bodies and Parkinson’s disease dementia, and vascular dementia. Twenty-six papers out of an initial 5,211 were reviewed and interpreted using a validated model of gait. Dementia was associated with gait characteristics grouped by slower pace, impaired rhythm, and increased variability compared to normal aging. Only four studies compared two or more dementia subtypes. People with AD are less impaired in pace, rhythm, and variability domains of gait compared to non-AD dementias. Results demonstrate the potential of gait as a clinical marker to discriminate between dementia subtypes. Larger studies using a more comprehensive battery of gait characteristics and better characterized dementia sub-types are required.
INTRODUCTION
Dementia is a growing global issue with 46.8 million people affected worldwide and numbers predicted to rise to 131.5 million by 2050 [1]. Dementia is identified by multiple cognitive impairments, which limit everyday functioning. It occurs predominantly in older adults and can be categorized into different subtypes. Alzheimer’s disease (AD) is the most common subtype, followed by Lewy body dementia (LBD) and vascular dementia (VaD) [2]. AD is characterized by gradual onset of memory impairment and is associated with neurofibrillary tangles and amyloid-β plaques contributing to neurodegeneration, particularly focal to the hippocampal region. Dementia with Lewy bodies (DLB) and Parkinson’s disease dementia (PDD) share symptomology and pathology; they have common key symptoms such as parkinsonism, cognitive fluctuations, visual hallucinations, and REM sleep behavior disorder, and are associated with Lewy body formation in the brainstem, limbic, and neocortical areas [3, 4]. Together, these dementia subtypes are referred to as LBD. LBD often has concurrent AD pathology, which alters clinical presentation. VaD is heterogeneous in nature and therefore cognitive changes vary greatly [5]. The most common findings are subcortical infarcts and white matter ischemia damaging frontostriatal circuits, leading to impaired attention, information processing, and executive function.
Misdiagnosis of dementia subtypes is problematic in AD and DLB; it is reported that 34–65% of cases are misdiagnosed [6], due to similarities in cognitive presentation and pathology. Some dementia cases, such as AD with subcortical infarcts, have mixed pathology which can further hinder accurate diagnosis. Often the need to distinguish subtypes is disregarded, as it is not thought to influence care. However, accurate diagnosis of DLB subtype is important to prevent mistaking cognitive fluctuations, key characteristics of DLB, as delirium; prevent inappropriate use of antipsychotics; and to facilitate early identification and treatment of motor symptoms, dysautonomia, falls, and other characteristic non-psychiatric symptoms [7]. Subtypes also have different prognoses, with DLB associated with more rapid decline and entering nursing care earlier [8]. The advent of disease modifying treatments will also necessitate subtype identification and dementia stratification for the optimal use of such therapies.
Diagnostic markers for dementia, such as cerebrospinal fluid, blood samples, brain pathology, and cognitive markers, are being investigated to distinguish dementia subtypes and improve accuracy of current clinical diagnoses [9]. More recently, gait (and its discrete characteristics) have been proposed as potential clinical biomarkers for dementia [10]. Gait is a complex skill requiring involvement from widespread brain regions (including those related to different cognitive functions, such as the frontal cortex and hippocampus). Changes in brain function can therefore lead to subtle changes in distinct gait characteristics, explaining why its features may be useful. Studies show a robust association between gait and cognitive function [11], and gait impairments precede and predict cognitive impairment and dementia [10, 12]. Therefore, evidence suggests quantitative gait analysis as a plausible diagnostic marker for early diagnosis of dementia. However, recent reviews have not addressed the role of gait to differentiate between dementia subtypes.
In consideration of this, the aims of this review are to establish quantitatively assessed gait differences between dementia and non-cognitively impaired older adults, review evidence for distinct gait profiles across dementia subtypes, and identify recommendations for future research. This review will focus on the most common subtypes of dementia: AD, VaD, and LBD (referring to DLB and PDD). This review will focus solely on single-task gait analysis as dual-task protocols (which involve walking while engaging in another task) vary widely in both methodology and type of secondary task (i.e., tests to assess different cognitive domains or manual function). Different tasks may produce different gait impairments and is therefore a subject for further detailed investigation beyond the scope of this review. Assessing differences in gait impairment during single-task walking is clinically useful, as it is a simple task to carry out and easy to understand—an important consideration for populations with cognitive impairment. For the purposes of this review, we will adopt a model of gait (Fig. 1) [16] as a framework to provide structure to the synthesis of literature and aid interpretation of data. We hypothesize that gait will be more impaired across multiple domains in dementia compared to controls, and that LBD and VaD will have reduced pace and increased variability when walking compared to AD, whereas AD will have more pronounced impairments in temporal characteristics of gait. Characteristics relating to reduced pace and increased variability are associated with impaired attention and executive function, while temporal characteristics of gait have been linked to memory [12].
Lord et al. [16]’s model of gait for older adults. Gait domains include pace, rhythm, variability, asymmetry, and postural control.
METHODS
Search strategy
Six databases were used for the search: Scopus, Embase, Web of Science, Psych Articles, Medline, and Psychinfo. Key terms for the search strategy are detailed in Fig. 2. The search was limited to papers published from 1946 to October 2016. Other eligible papers brought to the reviewers’ attention were also considered. Articles were included if they: 1) included at least one dementia subtype and control/other clinical cohort (i.e., Parkinson’s disease; PD) or two dementia subtypes or at least one dementia subtype at different stages of disease severity; 2) included quantitative gait characteristics, obtained from electronic gait analysis, wearable technology, motion capture analysis, or other suitable means; 3) were original articles; and iv) were written in English. Where an article included another clinical cohort (e.g., PD or mild cognitive impairment, MCI) or other clinical characteristics (e.g., urinary symptoms), only the data relating to dementia and gait was reviewed.
Flowchart of search strategy and extraction of eligible studies.
Data extraction
One reviewer (R.M.A.) screened the titles from the initial search and two reviewers (R.M.A. and B.G.) independently screened the abstracts to identify potential articles. Full-text articles were retrieved when reviewers could not determine the eligibility of the study from the title and abstract. All full-length articles were reviewed by three reviewers (R.M.A, R.M., and J.W). Data were extracted from eligible articles. The key characteristics of interest were: 1) dementia subtypes included, 2) gait parameters assessed, 3) method of gait analysis, and 4) main findings of the study with respect to gait. A quality assessment was conducted separately by two reviewers (R.M.A and J.W) and overall quality scores were determined for each study (see Supplementary Table 1).
Interpretation of data
Due to the wide and varying range of gait characteristics, several groups have proposed models of gait that categorize gait characteristics by domain using data reduction techniques [12 , 14–16]. Although comparable, there is no standardized model—different models emphasize different characteristics and domains. The model chosen for this review was validated in older adults and PD (see Fig. 1 for more details). Gait characteristics across studies were broadly mapped onto five core domains (Fig. 1; hypothesized to represent different neural networks involved in locomotor control) in order to structure data presentation and interpretation of results for within this review [11].
RESULTS
Search yield
The search strategy generated 11,515 papers after exclusion criteria were applied. After removing duplicates, 5,211 papers remained from the search (see Fig. 2). The initial title search yielded 376 papers with an abstract screening leaving 55 papers eligible for data extraction. Fourteen studies were excluded as they did not specify the subtype of dementia (n = 10), were not relevant to the review (n = 3) or had previously reported results in a paper included in the review (n = 1). Data were extracted from 42 papers. After data extraction, a further 16 papers were removed as they only reported timed gait speed or used functional tasks which required additional tasks, such as the Timed Up and Go test. All papers were published between 1983 and 2016.
Out of the remaining 26 articles, the majority of studies investigated AD (n = 25; 96%), followed by DLB (n = 2; 8%), PDD (n = 2; 8%), LBD (n = 1; 4%), VaD (n = 1; 4%), and unspecified non-AD dementia (n = 1; 4%). Two studies used PD for comparison, four used MCI, and 21 used older adult control groups.
Measurement of gait in dementia
Table 1 details the specific characteristics and findings for each of the reviewed papers. Quantitative gait analysis included the use of gait walkway systems [17 –25], accelerometers [26 –30], motion capture analysis systems [31 –35], pressurized foot-sensors [19 , 36–39], and combinations of these and other methods such as forceplates [40] and digital cameras [41]. One study did not define the instruments they used [42].
Descriptive information, methodology and main study findings of all cross-sectional studies. MMSE = Mini Mental State Examination; UPDRS = Unified Parkinson’s Disease Rating Scale; FAST = Functional Assessment Staging Test; CDR = Clinical Dementia Rating scale
To examine the wide range of reported gait parameters, all gait characteristics were mapped to one of the five domains of gait [16]. Commonly described gait parameters have been described in Supplementary Table 2. All 26 papers investigated pace [17 –42], 18 studies described characteristics relating to rhythm [18–20 , 41], 13 studies reported gait variability [17 , 39], two studies described characteristics of gait asymmetry [26, 27], and nine reported parameters relating to postural control [17–20 , 42].
Gait impairments in Alzheimer’s disease
25 studies assessed gait in AD [17–24 , 38–42]; 21 of these studies compared AD to controls [17–20 , 42], four studies compared AD to other dementia subtypes [18 , 42], four compared AD to MCI [22 , 26–28], and four studies compared AD severity levels [21 , 41].
In AD, all 25 studies assessed characteristics of pace, such as step velocity, step length, step, stance, and swing time variability [17–36 , 38–42] (See Table 2 for specific study details). People with AD typically walked with reduced pace [17–20 , 42] compared to controls, and were more impaired in severe AD [32, 36]. Reduced pace was also reported in AD compared to controls with low levels of white matter subcortical hyperintensities but not compared to controls with high levels of subcortical hyperintensities [20].
Recommendations for future research
In AD, 18 studies assessed characteristics of rhythm, such as step, swing, and stance time [18–20 , 41]. The majority found impaired rhythm in AD compared to controls [18 , 33–35]. One study found impaired rhythm with increased dementia severity [32]. One study found impaired rhythm in AD compared to controls with low levels of subcortical hyperintensities but not high levels [20].
In AD, 12 studies assessed features of variability, such as step velocity, step length, and step width variability [17 , 39]. Results were inconsistent between AD and controls; five studies found increased variability in AD [17 , 33], while four did not [24 , 35].
In AD, only two studies assessed features of asymmetry such as step time, swing, and stance asymmetry [26, 27]. Both compared AD to controls and MCI cohorts; no significant differences were found between any groups. In AD, nine studies assessed postural control of gait such as step width and step length asymmetry [17–19 , 42]. Typically, there were no significant differences between AD and controls for postural control characteristics of gait [17–19 , 43].
Gait impairments in Lewy body dementia
In LBD, three studies assessed gait. All studies assessed characteristics of pace [18 , 39] and generally found reduced pace compared to controls [18, 37]. Findings were also inconsistent between LBD and PD, with one study reporting reduced pace in LBD [39] and another study showing no group differences between PDD and PD [37]. No significant differences were found between subtypes of LBD [45]. In LBD, two studies assessed features of rhythm [18, 39] and found rhythm was impaired compared to controls [18]. One study reported impaired rhythm in LBD compared to PD but no significant differences between LBD subtypes [39]. In LBD, only one study assessed characteristics of variability [39]. It found no group differences between LBD and PD. The same study assessed postural control characteristics of gait in LBD and found no significant differences between controls and DLB. Asymmetry was not assessed in LBD.
Gait impairments in vascular dementia
One study assessed pace and postural control characteristics of gait in VaD [42]. It found reduced pace but no differences in postural control in VaD compared to both controls. Rhythm, variability, and asymmetry were not assessed in VaD.
Differences in gait between dementia subtypes and disease severity
People with AD demonstrated better pace compared to VaD [42]. In contrast, comparisons with LBD are inconsistent; one study found no difference in pace or rhythm between AD and DLB [18] while another reported reduced pace, impaired rhythm, and increased variability in LBD compared to AD [39]. One study compared mild and moderate severity AD to mild and moderate severity unspecified non-AD dementia [25]; for both severity levels, non-AD dementia had reduced pace and a larger stride width (a feature of postural control). However, impaired rhythm was only found in the non-AD group in the moderate cohort and impaired variability only in the non-AD group in the mild cohort. No significant differences for postural control characteristics were found between AD and VaD or AD and DLB [18, 42]. Surprisingly, no significant differences were found in pace or rhythm between AD and PD [39].
Reduced pace was reported with increasing dementia severity. All four studies comparing dementia severity found reductions in pace in the moderate-to-severe AD groups compared to the milder groups [21 , 41]. Results were inconsistent between AD and MCI; two studies reported slower pace in AD compared to MCI [27, 28] while two studies found no significant differences between these groups [22, 26]. No differences in characteristics of rhythm were found across dementia severity [22 , 41] and only one study reported impaired rhythm in AD compared to MCI [27]. Inconsistent results for variability were found between AD and MCI, with two studies showing increased variability in AD [26, 27] and two reporting no differences [22, 28]. One study found increased variability in moderate AD compared to mild AD [41] while another found increased variability in moderate and severe AD compared to controls; this was not found in mild AD [32]. Only one study found moderate AD had a larger stride width, a feature of postural control, compared to controls whereas mild AD did not [25]. No studies investigated asymmetry across dementia severity.
DISCUSSION
This review aimed to summarize available data on gait differences in people with dementia compared to controls and identify distinct gait profiles in dementia subtypes. This review clarifies previous findings of gait impairment in dementia compared to controls, specifically attributing impairments to pace and rhythm domains. However, we extend previous literature by identifying that dementia subtypes differ from each other in characteristics of pace, rhythm and variability, although the number of studies comparing subtypes (Fig. 3) and the range of gait characteristics described are limited.
Heat map detailing number of studies comparing groups. AD, Alzheimer’s disease; VaD, vascular dementia; DLB, dementia with Lewy bodies; PDD, Parkinson’s disease with dementia, LBD, Lewy body dementia; OD, unspecified non-AD dementias; MCI, mild cognitive impairment; PD, Parkinson’s disease.
Is gait in dementia distinct from normal aging?
Our findings provide insight into significant impairments in gait in AD, VaD, and LBD compared to non-cognitively impaired older adults that are consistent with our hypothesis. Reductions in pace was reported by the majority of studies; however, it was also the most commonly assessed characteristic. Other discrete gait characteristics may have identified key discrete differences and need to be assessed in order to develop distinct patterns of gait for dementia subtypes [11]. Temporal gait characteristics (i.e., those in the rhythm domain) appeared more impaired in dementia and were dependent on disease stage. Impairments in variability are inconclusive, largely due to inconsistencies in the variables measured.
Are gait impairments distinctive between dementia subtypes?
The findings of this review support the qualitative literature reporting that gait is more impaired in non-AD dementia subtypes compared to AD and emphasizes differences across pace, rhythm, and variability domains, which is somewhat consistent with our hypothesis [2]. Figure 4 provides a synopsis of the findings described. Only four studies compared gait across subtypes, highlighting a significant gap in the literature. Interestingly, no differences were found between PD and AD in one study—however, trends indicated that PD walked slower with a mean velocity of 1.13 meters per second and mean stride length of 115.82 centimeters compared to 1.2 and 125.33, respectively [39]. One study reported differences across MCI subtypes, which may relate to different dementia subtypes. For example, when compared to controls, amnestic MCI (aMCI) had slower pace, while non-amnestic MCI (naMCI) had slower pace and impaired rhythm [25]. This may be due to pathological differences with important implications, as aMCI usually develops into AD, while naMCI progresses into non-AD dementias, such as DLB or VaD [44]. Therefore, gait could act as an early marker to differentiate between dementia subtypes; however, further work is needed to determine this.
Associations between dementia subtypes and gait implied by the current literature, using Lord et al. [16]’s as a framework to interpret results.
Do gait impairments across dementia subtypes relate to cognitive impairments and their underlying neural correlates?
This review provides evidence for gait impairment in dementia subtypes reflecting cognitive impairments. Selective cognitive domains have been associated with discrete gait impairments which may reflect underlying pathology [12]. For example, characteristics of rhythm have been associated with memory, affected early in AD, while reduced pace and increased variability have been associated with impaired attention and executive function, affected early in LBD and VaD [11]. These cognitive impairments relate to the underlying neural correlates and pathological changes in different dementia subtypes. Our findings suggest that gait impairments may similarly reflect these differences. Dementias such as LBD have associated motor impairments due to disease pathology, such as neurodegeneration of the substantia nigra, which produces key motor impairments of which gait asymmetry and postural control may be a feature. It is worth noting, however, that despite these impairments, diagnosis in the early stages is still difficult. Therefore, while the differences in gait may not all be mediated by cognitive deficits and associated neural correlates, additional motor impairments may contribute to early differentiation.
An interesting question to ask is; do gait impairments reflect shared cognitive and pathological correlates consistent with different dementia subtypes? AD is associated with amnestic memory deficits predominantly due to amyloid deposition in the entorhinal cortex and hippocampus [45]. Atrophy of the hippocampus (involved in navigation and memory) is associated with decreased pace and variability [46], with speculative links between rhythm and the hippocampus; temporal aspects of gait have been associated with memory [12]. Reduced pace and increased variability are associated with frontal lobe atrophy and white matter hyper-intensities affecting frontal subcortical circuits in both dementia and older adults—areas that mediate attention and executive function [46, 47]. Frontal white matter lesions are key characteristics of VaD [5] and frontal neuronal loss is associated with LBD, lending explanation to pace and variability deficits. There are also correlations between increases in gait impairment with dementia severity and reduced frontal cerebral blood flow becoming more widespread [32], suggesting gait impairment is reflective of ongoing neural changes in dementia. However, the majority of research associating gait with specific brain regions focuses on gait speed; further research needs to be completed before drawing any conclusions in this area.
Limitations of current research and recommendations for the future
There are a number of discrepancies with the current research regarding quantitative gait assessment in dementia. Several additional studies using functional tasks (i.e., timed up and go) were identified but not included in this review, as they did not provide standardized measures of gait. This prevents comparison across studies and may be subject to confounding variables, such as impaired movement initiation. Of the studies that were included, distance walked, number of strides and steps, type of walk (i.e., continuous or intermittent), and gait analysis technique used (i.e., instrumented walkways, body worn sensors) varied. This limited interpretation when collating the results. Development of a standardized single-task gait protocol suitable for use in any clinic would be beneficial to aid generalizability of findings. This should include measuring at least 30 steps to assess variability characteristics [48]. Intermittent walks may be more suitable for dementia populations, particularly as the disease progresses, allowing for rest breaks as needed. Gait characteristics across studies also varied, with some studies limited to velocity and others assessing a wider range, such as stance time, step width, etc. Only two studies assessed features of asymmetry; this may be an oversight when considering dementias with notable asymmetric pathology, such as PDD, as asymmetric pathology may be reflected in gait outcomes. Studies should strive to assess a large range of spatial and temporal aspects of gait, to establish distinct gait profiles across dementia subtypes.
There was also a limited number of studies comparing dementia subtypes, as seen in (Fig. 3). The majority focused on differences between AD and controls, with only five studies investigating non-AD dementias. Although non-AD dementias such as LBD and VaD have notable gait impairments as described in the qualitative literature [2], quantitative gait assessment is needed to tease out subtle differences that may support diagnosis. More studies comparing subtypes are necessary. There were also discrepancies across studies regarding severity measures: a number of rating scales, such as the MMSE or the CDR, were used to establish stage of disease with inconsistent ratings determining disease stage. Studies were also restricted by small sample sizes and may not have provided a true picture of gait in dementia due to influence of outliers; studies should be adequately powered. Overall, the majority of studies were only of mediocre quality (see Supplementary Table 1 for more details). Therefore, we have provided key recommendations in Table 2 to guide future research.
Clinical implications
While gait impairments are recognizably present and often early markers of dementia subtypes such as VaD, PDD, or DLB [2], clinical recognition of gait deficits in AD is an emergent area of research. The National Institute of Neurological and Communicative Diseases and Stroke/Alzheimer’s Disease and Related Disorders Association (NINCDS-ADRDA) includes gait disturbances in their exclusion criteria for a diagnosis of AD [49, 50]. However, the findings from this review and previous qualitative studies show that gait impairments are more common in AD compared to controls [2]. Qualitative literature suggests that gait impairments are not present in mild AD; however, quantitative gait analysis reveals subtle discrete deficits in mild AD that progressively worsen. Equally, while parkinsonism is a core feature of DLB according to the latest diagnostic criteria [51], specific gait impairments have not been described, and the revised DLB criteria suggests that at least one clinical marker and a biomarker suggestive of LBD are necessary for early diagnosis. Although limited, the current evidence suggests that dementia subtypes have distinctive patterns of gait impairment. While more research is necessary in order to establish unique gait profiles in dementia subtypes, the end-result could complement current diagnostic criteria and show potential utility as a biomarker. Similar to acknowledging the specific cognitive domains impaired early in disease onset (e.g., episodic memory in AD), specific gait domains may also be impaired early (e.g., rhythm in AD). Changes in gait are also found prior to onset of cognitive decline; therefore, gait analysis at early intervals could contribute to early diagnosis of dementia. With advancing technology, quantitative gait analysis techniques are becoming smaller, portable, and more cost-effective and could prove a useful addition to a clinician’s toolbox.
Conclusion
Gait is impaired in dementia compared to cognitively intact older adults. Dementia subtypes may have discrete gait profiles but more research is necessary to establish these. Use of standardized protocols and assessment of a comprehensive range of spatiotemporal gait characteristics are necessary when studying gait in dementia and its subtypes. Future research should endeavor to establish quantitative gait analysis as a cost-effective and easily applicable clinical biomarker for dementia.
Footnotes
ACKNOWLEDGMENTS
This research was supported by the Alzheimer’s Society and the National Institute for Health Research (NIHR) Newcastle Biomedical Research Unit and Newcastle Biomedical Research Centre based at Newcastle upon Tyne Hospitals NHS Foundation Trust and Newcastle University. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health.
