Association of Gait Speed,Psychomotor Speed,and Dementia

INTRODUCTION

Slowness is one of the fundamental characteristics of aging [1] and is one of the main components of the decline in cognitive and physical functions. Slowness in physical function is usually measured by gait speed (GS) [2 –4], while slowness in cognitive function can be measured by processing speed (PS) such as time to complete Trail Making Test A (TMT-A) [5, 6], which is less effortful part of the test, than the other two processing test (TMT-B and the Isaac Set Test [7]) used in the Three-city (3C) Study [8]. Previous work conducted in the 3C Study showed that low GS and low PS were quite correlated [6] but not perfectly concordant [9], which could be explained by the fact that the two speed measures have different correlates (cognitive for PS and physical for GS). Moreover, both measures were independent predictors of mortality [9]. The 3C-Dijon and 3C-Bordeaux studies have already shown that GS is specifically associated with the risk of incident vascular dementia (VaD) either as single factor [10] or as a component of frailty [11]. On the other hand, low PS is recognized as one of the main characteristics of vascular cognitive disorders [10, 12]. Thus, vascular processes could be considered as a factor associated with age-related slowness. In turn, this may suggest that a common underlying mechanism explains the concomitant impairment in both GS and PS. In fact, GS and PS could be regarded as two different dimensions of age-related slowness; however, both performances were associated with higher risk of adverse health-related outcomes. If the vascular process is the major component of slowness, low GS and PS should be specifically associated with VaD. Alternatively, if another consequence of aging explained slowness, GS and PS could be related to either Alzheimer’s disease (AD) or other type of dementia, particularly those related to Lewy body/Parkinson’s disease.

The aim of the present study was to determine the association between GS, PS, and incident all-cause dementia (VaD, AD, or other dementia) among French community-dwelling older adults. To achieve these goals, we used the Bordeaux-3C cohort study data, a French longitudinal population-based cohort specifically designed to determine the risk of dementia and cognitive impairment attributable to cardiovascular risk factors.

METHODS

RESULTS

A total of 2,104 participants from the Bordeaux-3C center were initially included; however, subjects were excluded due to missing data on one of both GS and PS measures (n = 316), prevalent dementia (n = 22), specific conditions causing GS and PS impairment (arthritis, Parkinson syndrome, visual impairment, dependence, etc.) or incomplete measures of GS or PS (n = 423). We also excluded 78 participants, which were lost to follow-up. Analyses are based on 1,265 (60.1%) participants (Supplementary Figure 1, flow chart). Those excluded (n = 761) were older (p < 0.01) but did not differ on sex, level of education or dementia status at the end of follow-up. Mean age of the cohort was 74.0 years (SD 4.8), 60.4% were women, and 19.0% were positive for ApoE4. The descriptive characteristics of the sample (n = 1,265) per incident dementia events are presented in Table 1. A total of 203 incident cases of dementia (16.0%) were recorded during the 12-year follow-up period. Those who developed dementia were women (p < 0.001), older (p < 0.01), had more ApoE4 genotype (p < 0.01), had reduced GS (p < 0.001), and had reduced PS (p < 0.01).

Table 2 shows the incidence of dementia according to GS and PS. Cox proportional hazards models adjusted for sex, education, and ApoE genotype showed that for each SD decline in GS (Model 1) at baseline, the risk of developing dementia was 1.22 (95% CI = 1.09–1.36; p < 0.01). For each SD decline in PS (Model 2), the risk of developing dementia was 1.45 (95% CI = 1.29–1.63; p < 0.01). When GS and PS were included in the same model (Model 3), both were independent predictors of dementia development and the HRs remained almost unchanged (HR = 1.20, 95% CI = 1.07–1.34 for GS and HR = 1.44, 95% CI = 1.28–1.62 for PS). Considering GS and PS in the model after adjusted for gender, educational level, ApoE4, and Mini-Mental State Exam scores, similar results were obtained: the two variables were associated with increased risk of dementia (HR 1.220, 95% CI 1.089–1.366; p≤001 and HR 1.428, 95% CI 1.262–1.616; p≤0.01, respectively). When GS and PS were included in the same model, both were independent associated with dementia, and the HRs remained almost unchanged (Table 3).

In the analysis per dementia etiology (Table 4), for each decrease of one SD in either GS or PS, there was an increased risk of developing AD (HR = 1.40, 95% CI = 1.22 to 1.61; p < 0.01; HR = 1.16, 95% CI = 1.02 to 1.32; p = 0.02, respectively), or VaD (HR = 1.56, 95% CI 1.16 to 2.08; p = 0.003; HR = 1.34, 95% CI 1.05 to 1.71; p = 0.02, respectively). For the other dementias, only PS was a significant incidence predictor (HR = 1.55, 95% CI 1.17 to 2.07; p = 0.003) after adjusting for sex, educational level, and ApoE4 status. Only a non-significant trend was observed for the relation between GS and other dementias (HR = 1.29, 95% CI 0.99 to 1.69; p = 0.06). No significant interaction between GS and PS was observed.

DISCUSSION

In this population-based cohort study of older adults, we have demonstrated that both GS and PS were associated with an increased risk of developing AD, VaD, and other types (only PS) of dementia. Moreover, the effect size of these relationships was comparable for each subtype of dementia. The effect size for a one SD increase in PS (40% to 56% increase) was higher than for that of GS (16% to 34% increase). Nonetheless, there is overlap of the 95% CIs. Interestingly, when GS and PS were considered in the same statistical model as possible predictors of dementia, both were found to be independently related to incidence. Moreover, when considering GS and PS as covariates in the same model, the HRs of each measure did not change. This result suggests that GS and PS were independently associated with a risk of developing dementia over the 12-year follow-up period and that these associations were independent of age, sex, educational level, and ApoE4 genotype and cardiovascular risk factor (hypertension, diabetes, current smoking, and stroke). Our results confirm our previous findings showing that a decline in cognitive speed precede for many years the occurrence of dementia [21]. However, this paper added an important fact: the decline in cognitive or motor speed is not specific of a given disease, but predicts in a same manner all type of dementia. Thus, slowness in cognition and gait could characterize a vulnerable brain at high risk of dementia, even preceding the prodromal stage of neurodegenerative diseases [22, 23].

In the Swedish National Study of Ageing and Care in Kungsholmen, Welmer et al. [24] investigated the relation of walking speed, processing speed, and their changes over time with dementia incidence in older adults. Our results were only partially consistent with those reported in the Swedish National Study of Ageing and Care in Kungsholmen due to some differences: the characteristics of eligible patients; the PS evaluation, which was defined by a composite measure of standard tests (digit cancellation, trail making test-A, pattern comparison); and the author explored the relation both evolution of processing speed, walking speed, and incident of dementia. Therefore, by showing that the predictive value of each measure is independent, our results agree with the view that age-related slowness could not be explained using a single factor approach. Taken together, these results support the idea that GS and PS represent two different dimensions of aging and that both contribute to define subjects with at risk of developing dementia. To the best of our knowledge, the additive effect of GS and PS is a finding that has not been previously reported. We would like to emphasize that the instruments used to measure GS and PS in the present study are short and inexpensive, making them easily implementable in the clinical setting. Our findings support the use of PS and GS as reliable screening tools for older persons, independently of their characteristics and ApoE status.

The fact that GS and PS were associated with dementia of vascular etiology is not surprising. These findings are consistent with other studies that reported similar results [25 –27]. Our result may be partly explained by: 1) the role played by cardiovascular risk factors (such as hypertension, diabetes mellitus, etc.) with poor motor function and dementia in the elderly and its consequences on brain vessels [28, 29]; and 2) GS is significantly associated with muscle strength [30, 31] and muscle loss is highly correlated with inflammation, oxidative stress, and sex steroids levels which have been linked to cognitive decline such as VaD [32]. Moreover, the association between PS and vascular lesions in the brain has been already documented [33, 34].

By the same token, the association between GS, PS, and other dementias are not surprising since motor functions are early impaired in the majority of these diseases [35, 36].

On the other hand, the association between GS, PS, and AD is unexpected even if previous studies have found that slow GS is associated with AD incidence [37 –39]. Slow gait speed may constitute a risk factor for AD. Low GS possibly reflects an age-related reduction in physiologic reserve, which may render the brain more vulnerable to the accumulation of AD pathology and subsequent damage [40]. This result contributes to further understanding of the relation between cognitive and motor function in older adults [41]. Several studies [3 , 42–44] have shown that PS was associated to the risk of AD and the likelihood ratio test confirmed the incremental effect of informant reports in addition to the neuropsychological test scores. In AD, PS decrements are partly attributable to slower perceptuomotor, visual, and attention disorders.

Welmer et al. [24] also reported that cognitive deficits in the psychomotor speed domain may be expected to occur at very early stages of the dementia process, even before other clinical signs are present.

The strengths of the present study are the large number of participants selected in a community-based setting and the follow-up quality of the cohort. GS, PS, and global cognition were performed by trained psychologists. A selection bias could have overestimated our result. The number of missing values for PS was high, representing more than 10% of the sample. This illustrates the technical difficulties in obtaining this measure. Subjects with at least one error in the TMT-A were also excluded from the analysis [14]. But the number of missing values for GS and the number of exclusions for TMT-A was less frequent (less than 5%). The exclusion of subjects due to missing data and/or being lost to follow-up may have influenced our results, deriving in underestimating the relationship between both GS and PS and risk of dementia. Another limitation of our study is that we have not analyzed all the possible determinants of the two measures particularly coronary and renal disease, muscle weakness, previous stroke, and metabolic disease. In fact, as classically observed in epidemiologic studies, excluded participants had more co-morbidities and greater disability.