Clinical Applicability of Gait Analysis in the Early Detection of Cognitive Decline: A Cross-Sectional Study.

BACKGROUND Although promising results concerning dual-task gait predicting dementia progression in older adults with Mild Cognitive Impairment have been established, the question remains whether this can also be applied to older adults who are not yet suffering from overt cognitive impairments. Spatiotemporal gait analysis has proven to be valuable and reliable in this context. Nevertheless, these instruments are not common in clinical environments. This study aims to investigate whether easy applicable, inexpensive instruments can be used in the detection of gait alterations associated with early cognitive decrements. METHODS Eighty-two healthy community-dwelling older adults (age 60-89) with a MMSE ≥ 27 were enrolled in this study. Gait analyses using 10-meter walk test (10MWT) and GAITRite-instrument were performed under single and dual-task conditions. Performance on MMSE, CDT, MoCA, TMT-A, TMT-B, Digit Span Forward and Digit Span Backward were registered. Based on the tertiles of dual-task costs (DTC), subjects were categorized as “poor walkers” or “good walkers”. Comparative and predictive statistical analyses were conducted. walkers” performed signicantly worse on compared Spatiotemporal gait analysis poorer MMSECDT and MoCA spatial DTC showed lower ΔTMT verbal the dual-task during TMT-A Digit dual-task DTC identied using inexpensive tools, which are easy applicable in all clinical settings.

CONCLUSIONS Poor walking abilities can be associated with early cognitive decrements, even in a very early stage when cognitive performance is still more or less intact. Poor walking performance might therefore be considered a clinical precursor or warning signal for beginning cognitive decline. The subtle cognitive and gait decrements can be identi ed using inexpensive tools, which are easy applicable in all clinical settings.
Trial registration Not applicable Background During the past decades, the famous Latin quotation "Mens sana in corpore sano" has been the object of innumerable studies, even to the extent that it has literally been adopted to the titles of a bunch of scienti c research papers. At a time of persistent excessive social demands, stress and pressure to achieve, scientists became more than ever persuaded of the existence of an essential interplay between the (dys)function of one's body and one's mind (and by extension brains). Therapies and treatments have been adjusted accordingly. This assumption might even so create opportunities for early risk detection of mental or cognitive disorders in case of physical dysfunctions, or vice versa.
The use of dual-task paradigms has contributed a lot to the comprehension of human information processing and the cognitive-motor interrelationships. As such, it has been superseded that gait is a purely automated motor activity [1][2][3]. Nowadays, there is a general consensus that gait relies on higherlevel cognitive input [4]. Although there is some evidence that impaired executive functions may actually contribute to gait alterations [5], cause and effect has not been de nitively demonstrated [1]. Together with the fact that an altered gait pattern involves an increased risk of falling, this obscurity has inspired many researchers to compare cognitively healthy older adults with cognitively impaired older adults and assess the differences in gait and fall incidence in an attempt to identify high risk populations for falling.
Oppositely, Verghese et al. was one of the rst researchers to demonstrate that quantitative gait parameters can predict the development of dementia, in initially non-demented older adults [20,21].
These ndings were later corroborated by several researchers [22][23][24][25][26][27]. More than 10 years later, Verghese introduced a term "Motor Cognitive Risk Syndrome" which is characterized by a slow walking speed and subjective cognitive impairment and provides a clinical approach to identify individuals at high risk for dementia [28]. Recently, Montero-Odasso et al. conducted a study with the promising result that dual-task gait test may predict dementia progression in older adults with MCI [22]. However, Darweesh et al. commented that the question remains as to whether dual-task gait also predicts dementia among older individuals who do not yet have overt cognitive impairment [29]. In terms of optimal and early (preventive) interventional strategies, early detection of any clinical signs announcing or indicating beginning cognitive decline is crucial. At this preclinical stage, less advanced and low-cost treatments can indeed easily be implemented. Furthermore, the cognitively asymptomatic individual will have su cient cognitive abilities to understand and adequately execute training instructions. For the abovementioned reasons a cohort of healthy older adults without cognitive complaints seems an interesting and important target population.
Although sophisticated spatiotemporal gait analysis has been proven to be very sensitive and valuable in experimental settings such as academic research laboratories, these instruments are not common in clinical environments. This study therefore aims to investigate whether easy applicable, inexpensive instruments can be used in the detection of gait alterations in association with early cognitive decrements in healthy older adults.

Participants
Eighty-two community-dwelling older adults (age 60-89) were enrolled in this study. Participants were recruited through online advertising, yer distribution and by word of mouth. Older adults with a Mini-Mental State Examination (MMSE) score < 27 or indicating cognitive complaints were excluded since they do not belong to the population of interest of the current study. Also, major neurological or orthopedic disorders likely to directly impact gait, were not eligible. Older adults who were not able to walk independently with or without walking aids were excluded as well. This study was approved by the Ethical Committee of the Ghent University Hospital. Informed written consent was obtained from all participants.

Gait analyses
Spatiotemporal gait analysis was performed using the portable and reliable electronic GAITRite® walkway system (8.3 m x 0.89 m; CIR Systems Inc., Havertown, PA, USA) [30]. All subjects completed (i) one single task at a self-selected normal walking speed and two dual-task conditions: (ii) normal walking while enumerating animal names (verbal uency task) and (iii) normal walking while counting aloud backwards by 3's from 100 (arithmetic task). The three walking conditions were offered in a randomized order.
Bearing in mind its clinical practice, gait analysis was also performed in a more simple and inexpensive way. Whilst walking under the three conditions described above, time was registered over a distance of 10 meters (10-meter walk test, 10MWT). To eliminate acceleration and deceleration times, subjects started and ended walking 2 meter before and after the actual 10 meters distance. Walking speed was calculated by dividing 10 (meters) by the time registered.

Neuropsychological Assessment Battery
In order to gain more insight in the participants' cognitive status, several neuropsychological screenings were administered. Besides the well-known valid and valuable MMSE [31], Clock Drawing Test (CDT) was conducted as complementary assessment. CDT was scored on the four most reliable and practically useful items as proposed by Thalmann et al. [32]. The scores of the MMSE and CDT (maximum 30 and 7 respectively) were also combined to a single score (MMSECDT) with a maximum of 9 [32]. The Montreal Cognitive Assessment (MoCA) is a cognitive screening instrument that has been developed to screen patients who present with mild cognitive complaints, usually performing in the normal range on the MMSE [33]. Similar to the MMSE, it is scored out of 30 points, with higher scores re ecting better performance. Cognitive abilities that are examined by the MoCA are: visuospatial/executive function, naming, episodic memory, attention, language, abstraction, and orientation. In order to compensate for educational effects, Nasreddine et al. determined that one point should be added to the total MoCA score for individuals with ≤ 12 years of formal education [33]. Executive functioning, cognitive exibility and working memory were assessed by the Trail Making Test (TMT) [34]. In part A (TMT-A) the subject has to connect the randomly spread numbers from 1 to 25 in consecutive ascending order. In part B (TMT-B) alternating connection between randomly spread numbers (1 to 13) and letters (A to L) has to be made.
Participants were instructed to complete each part of the TMT as quickly and accurately as possible. In case of an error, the investigator immediately pointed it so that the subject could correct it and complete the test without errors (but at the expense of additional time) [35]. Time to complete TMT-A and TMT-B was recorded and represented their respective scores. The calculation of Delta-TMT (TMT-B -TMT-A) removes the potential bias due to the upper extremity motor speed element [35,36] and is suggested to be a more accurate measure of executive functioning than performance on TMT-B [37]. Finally, working memory was also assessed by use of the Digit Span Test. During the Digit Span Forward, the examiner enumerates digits which need to be recalled in the same order by the subjects. In the Digit Span Backward, series of digits have to be recalled by the subject in reverse order. The length of the series increases each time by one digit (respectively up to 9 and 8 digits). The test ends when the subject is not able to correctly recall the span in two consecutive attempts. The number of the longest span of correctly reproduced numbers represents the score of the tests. All questionnaires or measures described above are freely available and no license is required in order to administer them.

Statistical Analysis and calculations
All data were analyzed using SPSS version 26 for Windows (Statistical Package for the Social Sciences Inc., Chicago, IL, USA) and statistical signi cance was assumed at p < .05. Descriptive statistics were used to summarize the characteristics of the subjects.
As we were particularly interested in the extra effort induced by the dual-tasks, dual-task costs (DTC) of walking speed (10MWT), stride velocity, step time, step length, stride time and stride length were calculated as follows [38,39]: Based on the tertiles of the DTC's, subjects were categorized as 'poor walkers' (highest DTC tertile) or 'good walkers' (lowest DTC tertile). Subjects who ended up as 'intermediate walkers' (mid DTC tertile) were not used for comparative statistical analysis since we assumed that mediocre gait characteristics in this homogeneous healthy population would not be sensitive enough to reveal signi cant correlations with neuropsychological performance. Comparison of neuropsychological scores between 'poor walkers' and 'good walkers' was performed by the t-test for independent samples. However, if normality assumption was not ful lled or Levene's test for equal variances was not assumed, the non-parametric Mann-Whitney U test was used.
Separate linear regressions were used to test the cross-sectional association between dual-task gait costs as independent variables and neuropsychological outcomes as dependent variables. The DTC gait parameters and the associated neuropsychological tests from Tables 2 and 3 were entered in the regression models.

Subject characteristics
Characterization of the study population and mean neuropsychological outcomes are shown in Table 1. Mean age of the cohort was 71.1 years old and 40.2% were women. Lower education level was de ned as ≤ 12 years of formal education.
As we were interested in the potentially different impact of the two types of dual-task (verbal uency versus arithmetic), DTC's of the spatiotemporal gait parameters for both dual-tasks were calculated as explained earlier.
Comparative results for verbal uency DTC Categorization based on the DTC tertiles for the animal naming task resulted in signi cant different cognitive abilities when considering the walking speed DTC categorization of the 10MWT. Scores of CDT, MMSECDT, TMT-A and Digit Span Backward were signi cantly worse in "poor walkers" compared to "good walkers" ( Table 2).
Comparative results for arithmetic DTC Categorization based on the DTC tertiles of several spatiotemporal parameters for the serial subtraction task resulted in signi cantly different cognitive performance on some neuropsychological tests. 10MWT and spatiotemporal gait analysis revealed that 'poor walkers' (highest dual-task cost) signi cantly performed worse (p ≤ .05) on MMSE, CDT, MMSECDT, TMT-A, TMT-B, Delta TMT and some MoCA subscores (Table 3).
In order to ensure readability of Tables 2 and 3, it was opted only to include those neuropsychological test results that revealed signi cantly different scores between poor and good walkers.

Predictive value of DTC
To evaluate the potential predictive value of dual-task gait costs on neuropsychological outcomes, simple linear regression models were calculated for the entire study population.

Discussion
The results of this study suggest that poor walking abilities (based on high dual-task costs) can be associated with cognitive decrements, even in a very early stage when cognitive performance is still more or less intact. Apparently, these subtle cognitive decrements can be identi ed using inexpensive and timesaving gait analysis, applicable in all clinical settings. The 10-meter walk test appears to be a valuable instrument to discriminate between older adults with and without early onset in cognitive decline. The dual-task cost for both types of dual-tasks resulted in signi cantly different scores for CDT, MMSECDT, TMT-A and Digit Span Backward between poor and good walkers. Arithmetic dual-task cost could additionally discriminate performance on MoCA and the visuospatial/executive MoCA-subitem between poor and good walkers. Dual-task gait speed has recently been proven not to be a sensitive predictor of cognitive decline in healthy community-dwelling older adults [40]. However, the extra effort of dual-versus single task gait performance might be a more meaningful measure to use in this context.
Interestingly, spatiotemporal gait analysis using the GAITRite electronic walkway only revealed signi cant differences in cognitive performance when the most challenging dual-task (serial subtraction by threes) was performed. Since more demanding cognitive dual-tasks have been proven to induce higher dual-task costs [41], this may have resulted in stronger categorization in this trial. Besides, verbal uency mainly depends on semantic memory, whereas backward counting relies on the working memory. As both ambulation and arithmetic dual-tasking rely on the working memory, the increased competitive interaction with executive functions might explain this nding. Still, it is remarkable that categorization based on the 10MWT DTC's for verbal uency did yield signi cantly different cognitive performances. Gait velocity has repeatedly been proven to be a sensitive gait parameter for numerous outcomes [42,43]. The poorer cognitive outcomes for the poor walkers were also seen for GAITRite measured stride velocity DTC for verbal uency, although it lacked statistical signi cance. The nding that the 10MWT seemed to be more sensitive, might be attributed to the longer walking distance, eventually producing more reliable and stable gait patterns. Also, it might be expected that a larger sample size would result in signi cant neuropsychological differences for spatiotemporal gait verbal uency DTC's as well.
MoCA focuses on executive functions (working memory) and is therefore known to be a more sensitive instrument for identifying MCI than MMSE [44]. The fact that MoCA test results can cautiously be interpreted as a measure of working memory while backward counting also relies on (the same) working memory, might explain why some MoCA subitems appear to be signi cantly different when poor and good walkers were identi ed based on arithmetic DTC and not when they were identi ed based on verbal uency DTC. Interestingly, categorization based on DTC of spatial gait parameters (step length and stride length) results in signi cant differences in TMT-B, Delta TMT (assessing cognitive exibility) and the MoCA-language subscore, whereas categorization based on DTC of temporal gait parameters (stride velocity, step time and stride time) yields signi cant differences in MMSE, CDT, MMSECDT and the MoCAorientation subscore. Neuropsychological test results of poor and good walkers based on DTC-tertiles of temporal gait parameters are very analogous or even identical. Similarly, neuropsychological outcomes for poor and good walkers based on DTC-tertiles of spatial gait parameters are very comparable. Since the concerning gait parameters are closely related to each other, calculation of upper and lower tertiles will probably have generated the same subjects in the spatial or temporal gait related "poor" and "good" walker categories, which might explain this nding.
The orientation subitem of the MoCA test is also implemented in MMSE, which may explain that these neuropsychological test scores appear together in statistical analysis. Executive functions are known to play a critical role in gait regulation, especially when dual-task paradigms are introduced [1,34,45,46]. Dual-task gait abilities are even considered to be functional measures of executive functions [47]. Therefore, it might be surprising that distinctions based on dual-task costs could be made not only for TMT and CDT (measuring executive functions) but also for MMSE. However, some executive functions are important to memory implying that executive functions and memory are often not selectively impaired [48]. Besides, dual-task cost has previously been associated with episodic memory performance, con rming that other cognitive domains beyond executive function are involved in gait performance [41,[49][50][51]. This could contribute to the explanation of the nding that for spatiotemporal gait analysis not only tests speci cally investigating executive functions and working memory but also tests more focusing on cognition and memory (like MMSE) are signi cantly worse when dual-task costs are higher.
Regardless of the type of dual-task, the dual-task cost of 10MWT is able to predict performance on TMT-A and Digit Span Backward (both assessing attention, executive functioning and working memory) in a statistically signi cant way. As for prediction based on spatiotemporal gait analysis the arithmetic dualtask cost for stride velocity, step length and stride length was respectively able to predict performance on MMSE; MoCA-language; MoCA-language, TMT-B and delta TMT. The largest predictive value could be identi ed for MoCA-language subscore. Although the predictive values are rather poor, dual-task gait analysis seems to be a promising instrument in predicting early cognitive decrements.
As rightly commented by Darweesh et al. [29] most research in this area is targeting on a population that is known to be prone to develop dementia (i.e. MCI). However, similar to any pathological condition or disease, the preclinical stages of cognitive deterioration are critical and most susceptibale for early treatment or intervention strategies. These preclinical stages are characterized by executive dysfunction and working memory impairments [52]. Considering that these disturbances are very subtle in this period, the older individual generally is not aware of them since no discomfort during daily life activities is noticed. When the older adult does experience changes in daily functioning, the cognitive decline might already be drifted towards a more severe condition, requiring more advanced, expensive and challenging intervention strategies. For this study, we therefore deliberately opted to select a healthy older population not suffering from any subjective cognitive malfunction. For the same reason, clinicians should be encouraged to invest in systematic screening of neuropsychological functioning and walking abilities in the older adult, even if the latter does not experience functional problems so far. In practical terms, older individuals presenting with a poor performance on both neuropsychological screening and walking ability (high dual-task cost), as illustrated by the circular mark in Fig. 1, are suggested to be refered for further investigation as they are expected to be at risk for conversion towards more severe cognitive dysfunctioning.
The ndings in this report are subject to some potential methodological weaknesses. Although eighty-two older adults were enrolled, only upper and lower tertiles were used for categorization into "poor" and "good walkers", resulting in rather small groups to compare. However, for predictive analyses the entire sample was considered. Secondly, due to the non-prospective character of the study, it is impossible to conclude that the 'poor walking' subjects, associated with poorer neuropsychological performance, will actually develop cognitive disorders. Executive functions and attentional systems have been associated with prefrontal areas which undergo some structural brain changes during normal ageing [35,53,54]. The nding that higher dual-task costs represent poorer neuropsychological performance might therefore be not surprising and considered as normal or 'non-pathological'. However, it is generally accepted that healthy older adults exhibit some 'normal' strategies in response to dual-tasking, without widespread changes in the gait pattern [1]. We therefore consciously made categorizations based on several spatiotemporal gait parameters in this study, which all resulted in signi cantly worse neuropsychological performance. The subtle cognitive decrements of the poor walkers might thus be expected to be precursors of further cognitive deterioration, rather than to be ascribed to normal ageing.
Still, prospective study designs are needed to con rm this suggestion. Besides, it would be interesting to prospectively assess fall incidence within the identi ed categories based on walking and neuropsychological performances. For future research, it might be recommended to perform brain imaging (fMRI) to identify eventual changes in brain structure rather than using screening instruments only, since deviations in imaging might be determined earlier than its clinical manifestations that can be registered by screening instruments. However, in terms of implementation in clinical practice, it can be considered an advantage that simple neuropsychological screening instruments seem to be valuable alternatives that can be used in this context. Finally, in order to con rm and consolidate the results of this study, we agree with Darweesh et al. [29] that future research should de nitely focus on older adults without overt cognitive disabilities, rather than on older adults suffering from MCI (which is known to be a precursor for dementia development).

Conclusion
The aim of the current study was to investigate whether poor clinical walking abilities could be associated with early cognitive decrements in older adults without known cognitive complaints. This research shows indeed that the effort it takes to complete a cognitive dual-task while walking, can be related to early cognitive decrements in healthy older adults. Poor walking ability might thus be considered a clinical precursor or warning signal for beginning cognitive decline. Backward counting appeared to be the most sensitive type of dual-task in signalizing subtle cognitive decrements. Neuropsychological instruments, especially those assessing executive functioning and working memory, seem to be sensitive in case of high dual-task gait costs. To our knowledge, this is one of the few studies demonstrating that high dual-task gait costs can be related to early cognitive decrements, even in a healthy older population without overt cognitive impairment. Furthermore, this study shows that both cognitive and gait alterations are valuable when measured with inexpensive tools, which are easy applicable in all clinical settings. These ndings therefore suggest a potential role for clinicians in early detection strategies which are of utmost importance in the early initiation of interventional approaches to monitor eventual further cognitive deterioration in older adults. After all, individuals have a huge potential to reduce their dementia risk, even in later life [55].

Declarations
Ethics approval and consent to participate The study protocol was approved by the Ethical Committee of the Ghent University Hospital and all participants signed an informed consent.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Funding
This study was funded by the Special Research Fund, Bijzonder Onderzoeks Fonds (BOF), of the Ghent University [05v00608]. This grant had no role in study design, data collection, data analysis, data interpretation, or writing of the manuscript.
Authors' contributions DC and TRdM had the study idea. TRdM collected, analyzed and interpreted the data and was the major contributor in writing the manuscript. PD and DC interpreted the results and substantively revised the manuscript. All authors read and approved the nal manuscript. Tables   Table 1  Clinical characteristics of the subjects and neuropschological