Abstract
Background Cognitive impairments in Parkinson disease (PD) manifest as deficits in speed of processing, working memory, and executive function and attention abilities. The gait impairment in PD is well documented to include reduced speed, shortened step lengths, and increased step-to-step variability. However, there is a paucity of research examining the relationship between overground walking and cognitive performance in people with PD.
Objective This study sought to examine the relationship between both the mean and variability of gait spatiotemporal parameters and cognitive performance across a broad range of cognitive domains.
Design A cross-sectional design was used.
Methods Thirty-five participants with no dementia and diagnosed with idiopathic PD completed a battery of 12 cognitive tests that yielded 3 orthogonal factors: processing speed, working memory, and executive function and attention. Participants completed 10 trials of overground walking (single-task walking) and 5 trials of overground walking while counting backward by 3's (dual-task walking).
Results All gait measures were impaired by the dual task. Cognitive processing speed correlated with stride length and walking speed. Executive function correlated with step width variability. There were no significant associations with working memory. Regression models relating speed of processing to gait spatiotemporal variables revealed that including dual-task costs in the model significantly improved the fit of the model.
Limitations Participants with PD were tested only in the on-medication state.
Conclusions Different characteristics of gait are related to distinct types of cognitive processing, which may be differentially affected by dual-task walking due to the pathology of PD.
Between 25% and 80% of people with Parkinson disease (PD) develop cognitive impairments.1–3 Declines in cognitive performance are associated with the risk of mortality, declines in social interaction, and limited performance of instrumental activities of daily living.4 Moreover, declines in cognitive performance do not reliably respond to available pharmacological and surgical treatments for PD, suggesting that nondopaminergic pathways may be a contributing factor.5,6 Recent research also has suggested the involvement of nondopaminergic pathways in gait dysfunction in people with PD.7 Thus, examining the relationship between cognitive and gait performance in people with PD may provide further insight into shared neural substrates that may clarify the etiology, progression, and treatment of the disease. The purpose of this study was to examine the relationship between cognitive and gait performance in individuals with mild to moderate PD.
Most of the literature examining the relationship between cognition and gait performance in people with PD has focused on changes in gait performance while walking and completing a secondary cognitive task (dual task).8–11 Studies that examined walking while talking have demonstrated greater reductions in stride length and walking speed as the syntactic and phonological complexity of repeated sentences increased.8,9 Yogev and colleagues10 similarly found that gait speed decreased during a dual task (walking and counting backward) in both PD and healthy older adult control groups. However, gait variability significantly increased with a dual task in the PD group only.10 Yogev and colleagues10 suggested that a change in gait speed (and most likely stride length, as reported in previous studies8,9) is a normal protective response to dual tasking in both PD and healthy older adults and is related to increased attention demands during the dual task. Changes in variability, however, reflect the possibility that regulating stride-to-stride variability is no longer an automatic process in PD but has become an attention-demanding task.10
Age-related changes in gait speed and variability during simple overground walking (single-task walking) in healthy older adults challenge this theory. In well-functioning older adults, slower gait speeds during single-task walking are associated with decrements in executive function and memory.12,13 Beauchet and colleagues14 have shown that lower scores in information updating and monitoring are related to increased stride time variability during single-task walking among healthy older adults. This finding suggests that cognitive functions play an important role even when a person is walking without an additional cognitive load, potentially through overlapping neural systems. Yet, there is limited research examining the relationship between single-task walking and cognitive performance in people with PD. Moreover, cognitive impairments in PD manifest as deficits in speed of processing and working memory (ie, the simultaneous storage and manipulation of information) as well as executive function and attention abilities.15,16 Examination of a wider array of both gait and cognitive parameters across both single-task and dual-task walking conditions is needed to obtain greater insight into the relationship between cognitive and gait performance in people with PD.
In this study, we used correlation and regression analyses to compare gait spatiotemporal and variability measures and cognitive measures across a broad range of cognitive domains (processing speed, working memory, and executive function and attention) during single-task and dual-task walking conditions in people with mild to moderate PD. We hypothesized: (1) that decrements in both spatiotemporal gait parameters and gait variability would be associated with declines in all aspects of cognitive performance, executive function and attention, working memory, and speed of processing during single-task walking and (2) that the inclusion of dual-task costs on walking would strengthen these relationships.
Method
Participants
Thirty-five individuals diagnosed with idiopathic PD by a movement disorders–trained neurologist participated in the study. All participants were on a stable regimen of antiparkinsonian medication (>30 days) and were tested in the on-medication state. All participants reported they were fully responding to their medication at the time of testing. Participants with significant cognitive impairment (Mini-Mental State Examination [MMSE] score ≤24), a history of psychiatric disturbance (eg, major depressive disorder or generalized anxiety), falls (ie, a score >1 on the Unified Parkinson's Disease Rating Scale [UPDRS] item 13), cardiovascular disease, or severe unpredictable or painful episodes of motor fluctuation were excluded from the study. All participants were able to ambulate independently and did not wear orthotic or other corrective devices. Participant demographics are shown in Table 1. All participants signed an approved informed consent statement before the study.
Participant Demographicsa
Cognitive Assessment
Each participant completed a battery of cognitive tests to examine performance across domains of processing speed, visual and verbal working memory, and executive function and attention (Tab. 2). This battery of testing included traditional neuropsychological measures routinely assessed in individuals with PD. All cognitive tests were presented on a computer monitor in a fixed order to manage frustration associated with task difficulty. See Table 2 for descriptions of each task.17–22
Descriptive Statistics and Task Description for Cognitive Measures
Cognitive measures were entered into a principal components analysis using varimax rotation to obtain orthogonal factors suitable for use in linear regression. This analysis yielded 3 distinct factors (Tab. 2). For factor 1 (processing speed), response times in the star task, digit symbol substitution, Stroop color-word, Stroop index, and 0-back and 1-back tasks yielded the highest loadings. Higher scores indicate longer (ie, poorer) processing times. For factor 2 (working memory), accuracy on the operation span and both digit span tasks yielded the highest loadings. Higher scores indicate better working memory. For factor 3 (executive function and attention), Stroop XXX reaction time and accuracy on the 2-back, visual working memory, and digit symbol substitution yielded the highest loadings. Higher scores indicate better executive function and attention.
Gait Assessment
Participants completed 10 trials of barefoot overground walking along a 12-m walkway at their self-selected speed (single task) and 5 trials of barefoot overground walking while counting backward from a 3-digit number by 3's (dual task).10,23 Participants did not use assistive devices during the gait assessment and were given rest between trials when requested by the participant. Thirty-five retroreflective markers were placed over anatomical bony landmarks according to the Vicon Plug-in-Gait marker system (Vicon Nexus, Lake Forest, California). Kinematic data were collected at a sampling frequency of 120 Hz using a 10-camera motion capture system. Stride time, stride length, step width, swing time, and walking speed were calculated using standard definitions according to an algorithm programmed in Matlab (The Mathworks Inc, Natick, Massachusetts). The coefficient of variation (CV) (CV=SD/mean × 100%) was calculated to determine within-subject variability of stride time, stride length, step width, and swing time.24 To measure the change in gait due to a simultaneous cognitive task, dual-task cost was calculated as the log transformed (to normalize for skewed variables) ratio of dual-task to single-task value for each gait outcome measure.23
Data Analysis
A 1-way repeated-measures multivariate analysis of variance was completed to examine differences in gait between single- and dual-task walking. Bivariate Pearson correlation analyses described the strength of the association between cognitive factors and gait variables during the single-task walking condition. Correlations between cognitive factors and dual-task cost also were calculated to examine whether changes in gait due to dual-tasking effects are associated with cognitive performance. Multiple correlations were corrected with Bonferroni adjustments. Finally, to examine how dual-task walking affects the relationship between gait and cognitive performance, a block regression analysis was completed. To control for collinearity among the variables, the single-task gait spatiotemporal measures, single-task gait variability measures, and dual-task cost measures were each entered into separate principal components analyses using varimax rotation to obtain orthogonal factors suitable for use in the block regression. The single-task gait factors were entered first (model 1) followed by the dual-task cost factors (model 2) as predictors for each cognitive factor. The change in R2 values was evaluated. Separate regressions were completed for gait spatiotemporal and variability measures. Statistical analyses were performed using SPSS Statistics version 17.0 (SPSS Inc, Chicago, Illinois). Statistical significance was set at alpha <.05.
Role of the Funding Source
This study was supported by the National Institute of Aging (grant R21AG0033284-1A2).
Results
Associations Between Cognitive Factors and Gait Measures
Table 3 shows the bivariate Pearson correlations between cognitive factors, single-task gait spatiotemporal and variability measures, and dual-task cost. Results revealed significant moderate to strong (r≤.5) associations between the processing speed factor and stride length and walking speed. There were no significant associations between working memory or executive function/attention factors and gait spatiotemporal measures. Specifically, for the single-task condition, slower processing speed times were associated with shorter stride length and slower walking speed. A significant moderate to strong (r≤.5) negative association between the executive function and attention factor and step width variability was revealed. Thus, less accurate executive function/attention scores were associated with increased step width variability. There were no other significant associations between the processing speed or working memory factors and gait variability measures.
Correlations Between Cognitive Factors and Gait Parametersa
Single-Task Versus Dual-Task Walking
Participants walked slower with shorter steps while spending more time with both feet on the ground during the dual-task condition (Tab. 4). Furthermore, all dual-task costs document a pattern of reduced walking performance (Tab. 4). Interestingly, the analyses, when corrected for multiple comparisons, revealed a significant negative association between the processing speed factor and dual-task cost for only stride length. Given that the mean dual-task cost was negative, this finding would indicate that people with slower cognitive processing speeds experienced greater decrease in stride length in the dual task compared with people with faster processing speeds. No associations with dual-task cost were revealed for the working memory or executive function and attention factors.
Descriptive Statistics for Gait Spatiotemporal Measures, Gait Variability Measures, and Dual-Task Costa
Block regression models revealed that both gait spatiotemporal and variability factors are significant predictors of processing speed and executive function and attention, but not of working memory. Moreover, results from the block regression revealed a significant increase in variance accounted for when dual-task cost factors were entered into the model predicting processing speed. Specifically, using both gait spatiotemporal factors as predictors of the processing speed factor, model 1 (single task only) accounted for 27% of the variance in processing speed, and model 2 (single-task plus dual-task cost) accounted for 45% of the variance (with a significant change in R2 values as measured by the F ratio, P=.002) (Tab. 5). In the model assessing gait variability factors, model 1 (single task only) accounted for 24% of the variance in processing speed, while including dual-task cost factors (model 2) increased the variance accounted for to 34% (with a significant change in R2 values as measured by the F ratio, P=.04). Therefore, single-task gait spatiotemporal and variability factors account for 27% and 23%, respectively, of the variance in each model, and dual-task cost factors contributed an additional 28 and 10 percentage points, respectively (Tabs. 5 and 6). Therefore, the relationship between gait spatiotemporal and variability measures and processing speed was strengthened by inclusion of dual-task costs in the model.
Block Regression Model Predicting Cognitive Factors From Gait Spatiotemporal Parametersa
Block Regression Model for Cognitive Factors and Gait Variability Parametersa
Regarding the executive function/attention factor, entering both gait spatiotemporal factors accounted for 14% of the variance in model 1 (single task only). The addition of the dual-task cost factor in model 2 accounted for an additional 8 percentage points of the variance; however, this change was not significant (P=.06) (Tab. 5). For the model assessing gait variability factors, the inclusion of dual-task costs did not significantly change the amount of variance accounted for (21% in model 1 compared with 17% in model 2) (P=.85) (Tab. 6). This result suggests that the relationship between gait performance and executive function/attention in this group was not significantly affected by dual-task walking.
Discussion
This study was the first to examine the relationship between a large battery of cognitive tests and multiple gait spatiotemporal and variability measures in people with mild to moderate PD who are taking medication. Our results revealed that in people with mild to moderate PD in the on-medication state, stride length and gait speed were significantly associated with processing speed measures, and step width variability was significantly associated with executive function and attention measures. There were no associations with the working memory factor. Moreover, these associations were affected differently by dual-task walking. These results suggest that temporal and postural aspects of gait may be related to different cognitive domains, which can be differentially impaired by dual-task walking due to the pathology of PD.
Lord and colleagues25 have suggested that different control mechanisms may account for the temporal control of gait (stride time variability) and the postural control of gait (double stance time variability) in people with PD. Moreover, previous studies have reported that gait rhythmicity, rather than postural control, seems to improve preferentially with dopaminergic medication.26 In the current study, gait variables that were related to more temporal aspects of gait (ie, stride length and walking speed) were associated with processing speed. In contrast, gait variables related to postural aspects of gait (ie, step width variability) were associated with executive function/attention performance. These results are consistent with the previously proposed hypothesis that speed and temporal aspects of movement performance are controlled differently than postural aspects of movement performance.25 Furthermore, these findings suggest that these 2 dimensions of gait performance share overlapping systems or pathologies with dissociable cognitive processes, speed of processing, and executive function and attention, respectively.
Support for the hypothesis that temporal and postural components of gait performance are related to dissociable cognitive processes may come from neuroimaging studies. Processing speed has been consistently associated with white matter integrity in the frontal and parietal lobes.27–30 Both the frontal and parietal lobes have been implicated in the control of gait.31,32 In contrast, executive function has long been associated with activity in the prefrontal cortex in healthy adults and in people with PD.33,34 Cognitive tests specific to prefrontal function have been significantly associated with postural stability in patients with early to moderate PD.35 Thus, the associations between both stride length and walking speed and cognitive processing speed revealed in this study may potentially be accounted for by shared white matter networks connecting frontal and parietal regions, and the association between executive function and step width variability may be due to overlapping systems in the prefrontal cortex.
There were no significant associations in this study between the working memory factor and either single-task walking or dual-task costs. These results contrast with previous research demonstrating associations between working memory and gait performance in people with PD.36 This difference may be a reflection of the medication state of the participants tested. Previous research has shown that working memory deficits are improved with dopaminergic medication.37,38 Relationships between gait performance and working memory may not have been evident in this study because participants were tested on medication and were mildly to moderately affected by PD.
The results of this study extend the findings of previous studies on dual-task effects on gait in people with PD, which similarly have demonstrated reductions in stride length and gait speed as well as increased gait variability during dual-task walking in people with PD.8–11,23 We additionally found changes in step width and step width variability during dual-task walking. Although most gait spatiotemporal and variability measures changed significantly during dual-task walking, only the change in stride length (dual-task cost) was significantly associated with the processing speed factor. Moreover, both gait spatiotemporal and variability measures were significant predictors of the processing speed factor and the executive function and attention factor. When comparing the change in variability accounted for by each model (change in R2 values), only the models predicting the processing speed factor significantly improved when the model included dual-task cost. Thus, processing speed was related to single-task walking as well as to the amount of change in walking due to the dual task. In contrast, the change in variation accounted for by the models was not significant for the executive function and attention model. A recent study by Wild and colleagues11 revealed associations between executive function and attention measures and cognitive dual-task cost but not gait dual-task measures. Thus, changes in executive function and attention that occur in PD may affect single-task gait but do not disproportionately affect gait performance during the dual task. Taken together, these results suggest that the relationships between gait performance and processing speed and between gait performance and executive function and attention are differentially affected by dual tasking.
Increases in stride time and swing time variability during dual-task walking have previously been associated with a greater risk of falls in people with PD.23 Moreover, changes in step width and step width variability during walking have been associated with a higher risk of falls in healthy older adults.39–41 In keeping with the hypothesis provided by Yogev and colleagues,10 a change in gait speed during a dual task may indeed be a normal protective response to the increased attention demand of dual-task walking in people with PD. However, our results suggest that dual-task changes in gait speed may be accompanied by other undesirable changes in step width and gait variability even in people with mild to moderate PD who rarely fall (UPDRS item 13<1), which may potentially increase the risk for falling during dual-task walking. However, whether dual-task walking and falls are related is beyond the scope of this study, as fall history was not collected. Nonetheless, our results suggest that further research is needed regarding the relationship between dual-task walking and falls in people with PD.
Limitations
Limitations of this study include the lack of a control group and testing in the on-medication state. The associations revealed in this study cannot be fully attributed to PD but nonetheless provide positive insight into the relationship between gait and cognitive performance in people with PD. Because people with PD do not complete daily activities in the off-medication state, testing participants on medication provided information about how gait and cognitive performance are related in a participant's daily functioning state. However, future studies examining the effects of medication on the associations revealed in this study are needed. Due to the multiple cognitive tests and length of the gait tests, fatigue may have affected both cognitive and gait performance. However, participants were given rest between cognitive tests and gait trials when requested. Moreover, the gait tasks were not randomized. Thus, practice effects for the dual-task walking condition may potentially confound the results. Participants walked barefoot for the gait assessment to remove potential confounds of different supportive shoes, but this approach may confound the ecologic validity of the walking tests used in this study. Although the results of this study revealed 2 distinct associations between gait and cognitive performance, no imaging data were collected to verify the speculative brain regions or pathology involved.
In conclusion, the results of this study support the hypothesis that associations between cognitive and gait performance are dissociable and may be differentially affected by dual-task walking due to the pathology of PD. The slowing of overall processing speed may indicate a shared neural system underlying gait and cognitive performance that, when challenged by increasing cognitive load, affects both processes. In contrast, the association between executive function and attention performance and step width variability of gait may suggest the involvement of overlapping systems that require the prioritization of either cognitive or gait performance due to limited resources in other brain areas affected by PD. These divergent associations should be taken into consideration when treating people with PD and suggest the importance of evaluating both cognitive and gait performance in people with PD. Walking in an everyday environment requires some level of dual tasking. Therefore, strategies designed to increase global processing speed may benefit both gait and cognitive processes, ultimately improving ambulation throughout the environment. Moreover, those people with PD demonstrating declines in executive function and attention may benefit from rehabilitation strategies that target switching between gait and cognitive tasks to minimize the risk of falls.
The Bottom Line
What do we already know about this topic?
We know that people with Parkinson disease (PD) demonstrate both cognitive and walking impairments, but there is little information about the associations between these 2 impairments in PD.
What new information does this study offer?
New information from this study suggests that the slowing of overall processing speed may indicate a shared neural system underlying walking and cognitive performance. In contrast, the association between executive function/attention performance and step width variability of walking may suggest the involvement of overlapping systems that require the prioritization of either cognitive or walking performance.
If you're a patient, what might these findings mean for you?
These findings suggest that strategies designed to increase overall processing speed may benefit both walking and cognitive processes, and strategies that target switching between walking and cognitive tasks may minimize the risk of falling in people with PD.
Footnotes
Dr Stegemöller, Dr Okun, Dr Altmann, and Dr Hass provided concept/idea/research design. Dr Stegemöller, Dr Shelley, Dr Okun, Dr Altmann, and Dr Hass provided writing. Dr Stegemöller, Mr Wilson, Ms Hazamy, Dr Altmann, and Dr Hass provided data collection. All authors provided data analysis. Dr Stegemöller and Dr Hass provided project management. Dr Altmann and Dr Hass provided fund procurement. Dr Okun and Dr Hass provided study participants. Dr Okun, Dr Altmann, and Dr Hass provided facilities/equipment. Dr Stegemöller provided clerical support. Ms Hazamy provided consultation (including review of manuscript before submission).
This research was presented at the Annual Meeting of the American Society for Biomechanics; September 4–7, 2013; Omaha, Nebraska.
This study was approved by the University of Florida's Institutional Review Board.
This study was supported by the National Institute of Aging (grant R21AG0033284-1A2).
- Received June 16, 2013.
- Accepted February 14, 2014.
- © 2014 American Physical Therapy Association
References
Issue highlights
