Gait analysis in elderly adult patients with mild cognitive impairment and patients with mild Alzheimer’s disease: simple versus dual task: a preliminary report
Didier Maquet, PhD, Department of Motricity Sciences, University of Liege, ISEPK – B21 – Allée des Sports 4, B-4000 Liege, Belgium
Background/Aims: The aim of this study was to assess gait characteristics during simple and dual task in patients with mild cognitive impairment (MCI) and compare them with those of healthy elderly subjects and mild Alzheimer’s disease (AD) patients.
Methods: We proposed a gait analysis to appreciate walking (simple task and dual task) in 14 MCI, 14 controls and six AD subjects who walked at their preferred speed. A 20-second period of stabilized walking was used to calculated stride frequency, stride length, symmetry and regularity. Speed walking was measured by electrical photocells.
Results: Variables measured during simple and dual tasks showed an alteration of motor function as well in mild AD patients as in MCI patients.
Conclusion: At the end of this preliminary study, we defined a specific gait pattern for each cognitive profile. Further researches appear necessary to enlarge the study cohort.
Walking is one of the most universal of all human activities and the primary exercise recommended by public health authorities. Gait characteristics appear frequently altered in different pathologies. New tools enable the study of human locomotion (Andriacchi & Alexander, 2000). One of the most recent quantitative methods for analysing human gait is based on recording tri-axial accelerations, with a specific accelerometer, at a point close to the body centre of gravity (Auvinet et al., 2002). This method allows a fast and functional analysis of gait. Patients with Alzheimer’s disease (AD), the most common dementia, present greater risk of falls (Buchner & Larson, 1987; Morris et al., 1987). These patients generally have shorter step lengths, reduced gait speed, lower stepping frequency, and greater step-to-step variability than healthy elderly subjects (Visser, 1983; Buchner & Larson, 1987; Alexander et al., 1995; O’Keeffe et al., 1996). The specific nature of the dual task deficit in AD, despite the limited impact in healthy elderly subjects, offers relevant potential for developing walking protocols to aid diagnosis, to appreciate the evolution of this disease and to objectify the effectiveness of some treatments (Sala & Logie, 2001).
The transitional state between normal ageing and the early stages of AD has been designated as mild cognitive impairment (MCI) (Petersen et al., 2001). Patients with MCI suffer from impaired cognition, especially memory, to a greater extent than would be expected for a given age and education, but do not fulfil criteria for dementia (Morris, 1993). The progression rate to dementia is higher for patients with MCI compared with controls. However, the progression of this disease to AD remains uncertain. The aim of this study was to assess gait characteristics during simple and dual task in patients with MCI and compare them with those of healthy ageing (with no cognitive impairment) and mild AD patients.
Fourteen patients with MCI, 14 healthy control subjects and six patients with mild AD were included in the present study (Table 1).
Table 1. Demographic and morphological characteristics of healthy control subjects, mild cognitive impairment (MCI) patients and Alzheimer’s disease (AD) patients.
|Age (years)||74 ± 5||73 ± 4||74 ± 4|
|Height (cm)||165 ± 6||167 ± 10||168 ± 11|
|Weight (kg)||74 ± 10||67 ± 15||69 ± 12|
|Distance between great trochanter and external malleolus (cm)||76 ± 5||77 ± 7||79 ± 8|
All control subjects (to check that none of the healthy volunteers presented any memory impairments) and patients (AD and MCI) underwent a medical evaluation and a neuropsychological assessment.
The medical evaluation included an interview (to establish the subject’s full personal medical history) and a comprehensive clinical examination (height, weight and a cardiorespiratory, abdominal and above all neurological examination), the goal of which was to check for the absence of exclusion criteria. For the AD patients, family members and/or carers were questioned on living conditions, falls and medications; if required, we performed a telephone interview with the patient’s family doctor or consulting physician.
The evaluation performed by each subject included:
- 1 The Mini-Mental State Examination (MMSE) (Enzensberger et al., 1997) for orientation, learning, working memory (via mental arithmetic and word spelling), object naming, understanding simple instructions and copying a drawing. The 30-point score has a cognitive impairment cut-off at 24 (26 for subjects having received higher education).
- 2 The Mattis scale (Mattis, 1976) is another set of global cognitive evaluations which complements the MMSE. It investigates frontal and subcorticofrontal functions more broadly and can thus detect types of dementia other than AD. It explores attention, memory, verbal and motor initiation, conceptual abilities and visuoconstructive praxis. The cognitive impairment cut-off is 123 out of 144, which can be modulated depending on the subject’s education level.
- 3 A French version of the Grober and Buschke 16-item free recall/cued recall test (Van der Linden et al., 2004), which examines episodic memory.
- 4 Rey’s complex figure test (Rey et al., 1991) in order to evaluate visuoconstructive and visuospatial organization abilities. Scoring is both qualitative and quantitative (out of 36) and
- 5 The computerized alertness and divided attention sub-tests from the Test for Attentional Performance (TAP) battery (Zimmermann & Fimm, 2006).
Mild cognitive impairment patients had memory problems without any significant impairment in daily living activities or global cognitive deterioration. The patients had undergone neurological, neuropsychological and structural neuroimaging evaluations, and were classified as MCI since they had a Clinical Dementia Rating (CDR) score of 0·5 (Morris, 1993). They met the criteria for either amnestic MCI (with at least one memory test performance that were 1·5 standard deviations below the mean for age-matched controls) or multiple domain MCI (when cognitive performance was also decreased in another, non-memory domain) (Petersen et al., 2001; Winblad et al., 2004). Exclusion criteria were: mental retardation, less than four years of education, brain trauma, epilepsy, cancer, depression, any major systemic disease or any substance abuse. At the time of inclusion, all patients were free of medication that could noticeably affect brain function. All MCI patients had MMSE scores of minimum 24 arbitrary units (a.u.) at baseline evaluation (Folstein et al., 1975).
Patients with dementia were diagnosed as probable AD as defined by the National Institute of Neurological and Communicative Disorders and Stroke/Alzheimer Disease and Related Disorders Association criteria (McKhann et al., 1984). The exclusion criteria were the same as those used for MCI patients, and their CDR score reached 1. All AD patients had MMSE scores of 20 or over at baseline evaluation (Folstein et al., 1975), reflecting a mild stage of dementia.
The 14 healthy elderly adults (control subjects) were cognitively and neurologically intact. They lived in the community and were recruited by word of mouth. The exclusion criteria were the same as those used for the patients, and the NC had a zero CDR score.
All participants were 65 or older, with no story of fall or hospitalisation in the last 6 months. None of them had a history of musculo-skeletal disorder or orthopaedic problems. All of them were living at home.
The local Ethics Committee approved the study and all subjects gave their informed consent.
Gait analysis procedure
The gait analysis system used in this study (Locométrix™; Centaure-Metrix, Evry Cedex, France) includes an acceleration sensor, a recording device, and a computer program for processing the acceleration signals. The sensor is composed of two accelerometers placed perpendicularly to each other and housed in a moulded box (40 × 18 × 18 mm). The sensor is incorporated into a semi-elastic belt, which is fastened around the subject’s waist, so that the sensor is over the L3–L4 inter-vertebral space. One accelerometer is aligned with the median-lateral axis of the body, the other with the cranial–caudal axis. Signals from the sensor were recorded by a portable data logger with an acquisition frequency of 50 Hz. This device weighs 140 g and is housed in a box (65 × 22 × 12 mm). The recorded signals are transferred to a laptop computer using a transfer program operated under WINDOWS 98, formatted in files and analyzed by software developed in the MATLAB 5 environment (Mathworks, France).
The tests were carried out with subject walking at his/her comfortable speed down and back along a 45 m straight hospital corridor. No prompting signals were used. All subjects wore their usual walking shoes avoiding high heels or hard-soled shoes. Walking speed was measured by a system with electrical photocells developed in our University.
A period of steady state walking of 20·48 s was selected from the recording of each subject. This period contained about 1024 acceleration measurements and provided an optimal calculation time.
We calculated the following variables during simple and dual tasks:
- 1 Comfortable walking speed (m s−1).
- 2 Stride frequency (SF, in Hz): SF Movements of the left and right sides are identical from one cycle to the next in normal walking. Walking at a constant speed can, therefore, be considered to be the sum of a series of periodic stationary movements. Our analysis program used a Fast Fourier Transformation (FFT) to convert the cranial–caudal acceleration signal to the step frequency (the fundamental frequency of periodic movement). By definition, a complete stride includes two steps; thus SF corresponds to one half of the fundamental frequency and is expressed in Hz or stride per s.
- 3 Stride length (SL, in m) was calculated from the average speed (m s−1) divided by the SF (Hz).
- 4 Step symmetry (Sym, in arbitrary units) and stride regularity (Reg, in arbitrary units) were derived from two coefficients of correlation, C1 and C2, obtained by calculating the autocorrelation function of the vertical acceleration signal. We applied a Fischer Z transformation z(x) = 0·5 log [(1 + x)/(1−x)] to the coefficients C1 and C2 to obtain normal distributions (C1z and C2z). Since the symmetry index is the ratio of the two coefficients (C1z and C2z), we used a logarithmic transformation to obtain a normal distribution (Symtz LN). Stride symmetry describes the similarity of left and right cranial–caudal movements and is independent of fluctuations in the successive cranial–caudal movements of each limb. Reg describes the similarity of vertical movements over successive strides. Symmetry and regularity are dimensionless and
- 5 Number of stops during walking in simple and dual tasks.
Two walking conditions were randomized: simple task (walking) and dual task (walking with simultaneous backward counting). The stops and errors during dual task were measured.
Results are reported as mean ± SD. Since data were skewed, non-parametric statistics were used. Kruskal-Wallis ANOVA by ranks was used for multiple comparisons between the different groups. Categorical data were also analysed using Chi-square test. Spearman rank order correlation was used to explore possible relationships between gait parameters and some specific tests of cognitive function memory and attention. All the statistical analyses were performed using Statistical Analysis System 9·1 (SAS Institute Inc., Cary, NC, USA). A P-value of 0·05 was considered as statistically significant.
The demographic and morphological data on the subjects are given in Table 1. There was no statistical difference between the three groups. The same sex ratio was observed (50% males and 50% females) for the three groups.
Table 2. Comparison of gait variables between three groups during simple task.
|Walking speed (m s−1)||1·4 ± 0·13a||1·22 ± 0·15a,b||1·02 ± 0·36b|
|Stride frequency (Hz)||1 ± 0·08a||0·9 ± 0·05b||0·95 ± 0·17a,b|
|Stride length (m)||1·41 ± 0·1a||1·36 ± 0·13a,b||1·13 ± 0·45b|
|Symmetry (a.u.)||202 ± 31a||224 ± 25a||209 ± 77a|
|Regularity (a.u.)||276 ± 35a||287 ± 29a||227 ± 82b|
|Stops (a.u.)||0 ± 0a||0 ± 0a||0 ± 0a|
All subjects performed the walking test in simple task conditions without stop. Walking speed, SL and Reg were significantly lower in AD patients in comparison with control subjects and MCI patients. In contrast, SF was significantly reduced in MCI patients compared to healthy controls.
Table 3. Comparison of gait variables between three groups during dual task.
|Walking speed (m s−1)||1·3 ± 0·14a||1·05 ± 0·21b||0·74 ± 0·26c|
|Stride frequency (Hz)||0·94 ± 0·07a||0·81 ± 0·13b||0·81 ± 0·23a,b|
|Stride length (m)||1·38 ± 0·15a||1·3 ± 0·12a||1 ± 0·42b|
|Symmetry (a.u.)||250 ± 35a||216 ± 21a||206 ± 61a|
|Regularity (a.u.)||258 ± 38a||224 ± 47a||206 ± 61b|
|Stops (a.u.)||0 ± 0a||0·07 ± 0·27a||0·17 ± 0·41a|
|Errors (a.u.)||0·5 ± 1·16a||0·36 ± 1·08a||2·16 ± 3·92b|
During the dual task, the walking speed appeared significantly different between the three groups. Walking speed was the lowest in AD patients. None stop was observed in controls in contrast with some MCI and AD patients.
Comparison of gait analysis during simple and dual tasks
All control subjects performed the two walking conditions without stop. In these subjects, changes of walking speed and SF (dual versus simple task) were significant (Table 4).
Table 4. Comparison of gait variables during simple versus dual task.
| Walking speed (m s−1)||1·4 ± 0·13||1·3 ± 0·14||<0·05|
| Stride frequency (Hz)||1 ± 0·08||0·94 ± 0·07||<0·05|
| Stride length (m)||1·41 ± 0·1||1·38 ± 0·15||ns|
| Symmetry (a.u.)||202 ± 31||250 ± 35||ns|
| Regularity (a.u.)||276 ± 35||258 ± 38||ns|
| Stops (a.u.)||0 ± 0||0 ± 0||ns|
| Walking speed (m s−1)||1·22 ± 0·15||1·05 ± 0·21||<0·05|
| Stride frequency (Hz)||0·9 ± 0·05||0·81 ± 0·13||<0·05|
| Stride length (m)||1·36 ± 0·13||1·3 ± 0·12||<0·05|
| Symmetry (a.u.)||224 ± 25||216 ± 21||<0·05|
| Regularity (a.u.)||287 ± 29||224 ± 47||ns|
| Stops (a.u.)||0 ± 0||0·07 ± 0·27||ns|
| Walking speed (m s−1)||1·02 ± 0·36||0·74 ± 0·26||ns|
| Stride frequency (Hz)||0·95 ± 0·17||0·81 ± 0·23||<0·05|
| Stride length (m)||1·13 ± 0·45||1 ± 0·42||ns|
| Symmetry (a.u.)||209 ± 77||206 ± 61||ns|
| Regularity (a.u.)||227 ± 82||206 ± 61||ns|
| Stops (a.u.)||0 ± 0||0·17 ± 0·41||ns|
In contrast, during dual task, results showed that the Sym was significantly better in healthy controls than in both patient groups.
In the MCI group, changes of walking speed, SF, SL and Reg (dual versus simple task) were significant (Table 4).
Results of AD patients showed that changes of SF (dual versus simple task) were significant (Table 4).
Correlations between gait variables and neuropsychological variables
In MCI patients, significant correlations between gait variables and neuropsychological variables were reported (Table 5).
Table 5. Correlations between neuropsychological variables and gait variables in mild cognitive impairment (MCI) patients.
|Speed||r = 0·51*||r = 0·55*||r = 0·40||r = 0·48||r = 0·40||r = 0·36||r = −0·66*||r = −0·27*|
|Stride length||r = 0·59*||r = 0·65*||r = 0·15||r = 0·39||r = 0·04||r = 0·25||r = −0·36*||r = −0·22*|
|Stride frequency||r = 0·28||r = 0·28||r = 0·57*||r = 0·27*||r = 0·63*||r = 0·35*||r = −0·78||r = −0·19|
|Regularity||r = 0·42||r = 0·20||r = 0·17||r = −0·02||r = 0·01||r = −0·05||r = −0·47||r = −0·17|
|Symmetry||r = 0·46||r = −0·20||r = 0·32||r = −0·24||r = 0·36||r = −0·36||r = −0·48||r = 0·40|
|Symmetry (dual task versus simple task)||r = −0·45||r = −0·41||r = −0·51||r = 0·58*|
The subjects’ ability to perform more than one task at a time has been a topic in the literature on human attention and memory. Furthermore, age-related changes in gait parameters have been associated with an increased risk of falling and an adaptation to a safer gait (Maki, 1997). Our results showed that changes of walking speed and SF (dual versus simple task) were significant in the healthy elderly subjects. Hollman et al. (2007) reported a decrease in walking speed and an increase of step-to-step variability under a dual task condition in elderly subjects. Beauchet et al. (2003) showed also that cognitive activity during walking reduces gait velocity and increases variability in stride velocity in well-elderly women. Walking is, in normal conditions, a highly automated and rhythmic motor behaviour that is mostly controlled by subcortical locomotor brain regions (Nutt et al., 1993). In elderly subjects, the dual task interfered with gait parameters suggesting the involvement of higher cortical regions for the motor control of gait in this population (Beauchet et al., 2003). Furthermore, walking requires more attention and thus more cortical involvement with increasing age (Woollacott & Shumway-Cook, 2002).
We reported an intriguing finding with regard to the improvement of Sym during dual task walking in healthy elderly subjects. Sym describes the similarity of left and right cranio-caudal movements. This parameter was determined by calculating the two coefficients of correlation obtained by measuring the autocorrelation function of the vertical acceleration signal. Further researches are needed to confirm this observation that might be related to the rhythmic stimulation produced during backward counting in healthy subjects. Accordingly, Thaut et al. (1996) and Enzensberger et al. (1997) have reported the positive effect of metronome stimulation during walking in patients with Parkinson disease.
Our results in AD patients confirmed data previously published in literature (Visser, 1983; Buchner & Larson, 1987; Alexander et al., 1995; O’Keeffe et al., 1996). We showed that AD patients present shorter SL, reduced walking speed and lower Reg, particularly during dual task conditions. Furthermore, AD patients presented a significant increase of errors during backward counting. Cocchini et al. (2004) reported that dual task effects of walking (when talking in AD patients) display a deficit of executive functions in these patients. We observed no significant difference in errors measured during backward counting between MCI and control groups.
Our results demonstrated that SF during simple and dual tasks conditions appears significantly lower in MCI patients than in healthy elderly subjects. Furthermore, during dual task walking, MCI patients showed a walking speed significantly different, not only from normal control subjects but also from AD patients. These findings strongly suggest that the gait analysis protocol proposed in our study could contribute to distinguish the performances of the healthy elderly subjects, MCI patients and AD patients.
Conversely, Pettersson et al. (2005) reported that motor function was affected in very mild AD but not in MCI patients. Such difference might be caused by the experimental protocol proposed to assess muscle function. These authors used clinical performance-based tests: Bergs Balance Scale (Berg et al., 1992a,b; Berg et al., 1995), Falls Efficacy Scale (Tinetti et al., 1990; Hellstrom et al., 2002), Timed up and Go (Podsiadlo & Richardson, 1991), talking while walking with a measure of time difference between the two conditions and Tinetti gait protocol (Tinetti, 1986). While our data recording remains simple and fast in the clinical use, we proposed, in the present study, a gait analysis with more quantitative measurements.
This study also explored the possible relationship between gait parameters and neuropsychological tests (dedicated to cognitive function, memory and attention) that all control subjects and patients had underwent before the gait analysis. A positive correlation (P<0·05) was observed between walking speed and total MMSE score in MCI patients which suggests that the higher the global cognitive level, the faster the gait. Furthermore, the SF was positively correlated with the overall Mattis score and the Mattis initiation subscore. This seems to imply that the higher the overall cognitive performance level, the higher the SF. Gait speed and SF thus appear to be related to global cognitive function. Stride frequency is also related to executive function.
Moreover, change in stride symmetry between simple and dual tasks is positively correlated with the lengthening of the reaction time in divided attention sub-tests included in the attentional performance battery. Hence, the greater the difficulty subjects have in sharing their attentional resources between two tasks (i.e., visual and auditory tasks), the higher the coefficient of variation of symmetry and the stronger the decrease in symmetry when moving to a dual task. These results prompt a hypothesis whereby gait symmetry is not automatic but, in fact, requires attentional resources.
The main study limitation we faced is related to the limited number of AD patients. Our study cohort appears too small for definite conclusions. However, the detection of gait abnormalities in MCI and AD patients seems essential because different researches have shown that impaired motor function may restrict activities, induce risk of falls and even predict death (Murphy et al., 2002; Guralnik & Ferrucci, 2003).
The aim of this preliminary study was to propose an original tool for the evaluation of gait parameters. We demonstrate that this protocol is able to define a specific gait pattern for each cognitive profile. Further researches appear necessary and could contribute to identify one or several gait parameters in MCI patients, which could serve as predictive clinical factors of progression towards dementia in this population.