https://orcid.org/0000-0001-5220-4927
Abstract
Gait variability is a marker of cognitive decline. However, there is limited understanding of the cortical regions associated with gait variability. We examined associations between regional cortical thickness and gait variability in a population-based sample of older people without dementia.
Participants (n = 350, mean age 71.9 ± 7.1) were randomly selected from the electoral roll. Variability in step time, step length, step width, and double support time (DST) were calculated as the standard deviation of each measure, obtained from the GAITRite walkway. Magnetic resonance imaging (MRI) scans were processed through FreeSurfer to obtain cortical thickness of 68 regions. Bayesian regression was used to determine regional associations of mean cortical thickness and thickness ratio (regional thickness/overall mean thickness) with gait variability.
Smaller global cortical thickness was only associated with greater step width and step time variability. Smaller mean thickness in widespread regions important for sensory, cognitive, and motor functions were associated with greater step width and step time variability. In contrast, smaller thickness in a few frontal and temporal regions were associated with DST variability and the right cuneus was associated with step length variability. Smaller thickness ratio in frontal and temporal regions important for motor planning, execution, and sensory function and greater thickness ratio in the anterior cingulate was associated with greater variability in all measures.
Examining individual cortical regions is important in understanding the relationship between gray matter and gait variability. Cortical thickness ratio highlights that smaller regional thickness relative to global thickness may be important for the consistency of gait.
Intra-individual gait variability is the fluctuation of a gait measure from one step to the next (1). Gait variability is greater with advancing age (2), and in older people, greater variability is associated with adverse health outcomes such as cognitive decline, disability, falls, and frailty (3–6). Although it is understood that poorer cognitive (7) and sensorimotor (8) functions are associated with greater variability, there is limited understanding as to whether age- or disease-related changes in brain structure may contribute to the amount of variability in gait.
Of the few studies undertaken to date, most have focused on cerebral vascular lesions and gait variability. A higher burden of white matter hyperintensities or brain infarcts is associated with greater variability in both spatial and temporal measures (9–11). Findings have been mixed from the few studies examining the link between gray matter and gait variability, with associations reported between lower parietal volume and higher stride time variability (12), greater hippocampal volume and higher stride time variability (13,14) or no association between total (15) or hippocampal volume (13,16) and variability in frontal or sagittal spatial gait measures. Inconsistent findings might be due to the different populations that have been studied (ie, from memory clinics (13,14) or community living older people (12,15,16)), small sample sizes (12,13,16) or different a priori regions of interest (13,14,16).
Gray matter volume is the product of cortical thickness by surface area (17) and varies with total intracranial volume (18). In contrast, cortical thickness is not associated with total intracranial volume (18) and is more sensitive to age-associated changes in some regions such as the parietal lobe (17). However, few studies have examined associations of cortical thickness with gait (19,20). In people with cerebral small vessel disease, smaller cortical thickness in widespread areas including frontal, inferior partial, and superior temporal regions were associated with slow gait speed, shortened stride length, and broader stride width (20). In people with subcortical vascular cognitive impairment, gait disturbances assessed visually were associated with smaller thickness in multiple brain regions (19). However, no studies to the best of our knowledge have examined associations of cortical thickness with gait variability. Examining individual gait variability measures is important as these measures are not homogenous, loading onto different aspects of gait control (ie, step time variability loads onto pace, whereas step width variability loads onto balance control (21)).
Therefore, the aim of this study was to examine in a population-based sample of older people without dementia, the associations between regional mean cortical thickness and individual gait variability measures. Furthermore, some regions of the cortex show greater thinning rates per decade relative to others (22). To determine whether regional thickness relative to global cortical thickness was associated with gait variability, a thickness ratio measure (regional thickness/total mean thickness) was also examined.
Method
Study Participants
Older adults aged 60–85 years were randomly selected from the Southern Tasmanian electoral roll into the Tasmanian Study of Cognition and Gait (TASCOG). Participants were excluded if they were residing in a residential aged care facility or had dementia or a history of Parkinson’s disease. Participants were additionally excluded if they had any contraindications to magnetic resonance imaging (MRI). Ethical approval was obtained from the Sothern Tasmanian Health and Medical Human Research Ethics Committee (ethics approval no. H7947). Written consent was obtained from all participants.
Gait Assessment
Gait measures were assessed using the footfalls recorded on a computerized 4.6-meter GAITRite walkway (GAITRite system, CIR Systems, PA). Each participant completed six walks at their preferred pace. A walk consisted of an additional 2 meters before and 2 meters after the walkway to allow for steady pace walking. Gait speed was obtained from the GAITRite software. We calculated intra-individual variability of four gait measures to represent (a) both spatial and temporal measures of gait in sagittal and coronal planes (2) and (b) different aspects of gait control (21). Variability for each measure was calculated as the standard deviation of each measure averaged across all steps of the six walks. These gait variability measures have previously been associated with risk of multiple falls and dementia (23). Though GAITRite offers several gait measures, we did not include measures such as single support time or swing time as they were highly correlated with some of the other selected gait measures (ie, step time).
MRI Acquisition and Processing
All MRI scans were obtained from a single 1.5-Telsa General Electric scanner (LX Horizon, General Electric, Milwaukee, WI). The standard MRI protocol of MRI acquisition has been previously described (24). Mean thickness of the 68 cortical regions of the Desikan-Killiany atlas (25) and each hemisphere was estimated using FreeSurfer 5.3. FreeSurfer is an automated pipeline that incorporates correction for magnetic resonance field bias non-uniformity, removal of non-brain tissue, registration to volumetric and spherical atlases and, segmentation of cortical and subcortical structures (26,27). Segmentation errors caused by white matter hyperintensities were automatically corrected using information from co-registered fluid-attenuated inversion recovery (FLAIR) images. Results were inspected and images with poor quality were excluded from analysis. Mean thickness of the entire cortex was calculated by averaging mean thickness for each hemisphere provided by FreeSurfer. Regional mean thickness was obtained from FreeSurfer for each region, and the thickness ratio was calculated by dividing regional mean thickness by mean thickness of the entire cortex.
Cognitive Assessment
Cognitive function was assessed with a battery of neuropsychological tests, including (a) Executive function: using the Victoria Stroop test, the Controlled Word Association Test (using the letters F, A, and S). (b) Processing speed-attention: using the Digit Span and Digit Symbol Coding and the Symbol Search subsets of the Wechsler Adult Intelligence Scale-III, (c) Visuospatial function: using the Rey Complex Figure copy task, (d) Memory: using the Hopkins Verbal Learning Test—Revised and the delayed reproduction of the Rey Complex Figure copy task. Based on test scores, two clinical neuropsychologists reached consensus on a classification of no cognitive impairment if a participant scored >1.5 SD of age, sex, and education appropriate norms in all tests under each domain.
Other Measures
Demographics and medical history were recorded with a self-reported questionnaire. Medical history included hypertension, hypercholesterolemia, ischemic heart disease, diabetes mellitus, stroke, and lower limb arthritis. Height was measured with a Leicester height stadiometer.
Data Analysis
The associations between the mean thickness of the entire cortex and each gait measure were examined using Bayesian regression models adjusting for age, sex, and height (correlated with gait and brain size). For regional associations between cortical thickness and gait, we followed the approach of Peterson (2003) and Thompson (2007) (28,29), using a multilevel model. All 68 region measures were analyzed in a single model. Correlations between repeated measures (regions from the same individual) were modeled using a random intercept term for each individual. Bayesian regression approaches implemented in the R package “brms” (30,31) were used to fit models and perform inference. This approach was chosen over traditional multilevel regression as Bayesian regression does not constrain the final model distribution to be normal (our traditional model residuals were skewed). In addition, it did not exhibit problems with optimization parameters for which model solutions would converge. Models predicted thickness (or thickness ratio) as a function of gait variability and cortical regions. Terms for gait variability, and the interaction between region and gait variability, were included in these models to allow each region to have its own effect of gait variability on thickness. The sum of these two terms was reported for each region as the association between thickness and gait variability. A negative association for a region indicates a smaller cortical thickness (or thickness ratio) is associated with greater gait variability, whereas a positive association for a region indicates greater cortical thickness (or thickness ratio) is associated with greater variability. We used the credibility intervals as the index for hypothesis testing. The regions that do not cross zero are reported. Additionally, as a measure of the strength of associations, one-sided Bayes factors are reported.
Two secondary analyses were performed. Firstly, associations between regional cortical thickness measures and gait speed were explored. A positive association between a region and gait speed indicates smaller cortical thickness with slower gait speed, whereas a negative association indicates greater cortical thickness with slower speed. Finally, associations between regional cortical thickness measures and each gait variable were also examined in people with no cognitive impairment on neuropsychological testing.
Results
From the initial sample (n = 425), participants with missing gait data (n = 9), inadequate understanding of English (n = 1), and difficulty walking without an aid (n = 6) were excluded leaving 409 participants. Additionally, those with missing (n = 48) and poor-quality imaging data (n = 11) were excluded, leaving 350 participants for the final analyses. Participant characteristics are summarized in Table 1. The mean age of participants was 71.9 ± 7.1 years and 56% were male. Smaller overall mean cortical thickness was associated with greater variability in step width (β −0.10, 95% CI −0.19, −0.00) and step time (β −0.07, 95% CI −0.12, −0.01), but not with double support time (DST) variability (β −0.02, 95% CI −0.09, 0.04), step length variability (β −0.07, 95% CI −0.17, 0.02) or gait speed (β 0.07, 95% CI −0.04, 0.16).
Characteristics of Participants (n = 350)
| Variable | ||
|---|---|---|
| Age (y), mean, SD | 71.9 | 7.1 |
| Male, n, % | 196 | 56.0 |
| Greater than high school education, n, % | 159 | 45.3 |
| BMI (kg/m2), mean, SD | 27.6 | 4.2 |
| Self-reported medical history, n, % | ||
| Hypertension | 174 | 49.7 |
| High cholesterol | 142 | 40.6 |
| Angina | 45 | 12.9 |
| Myocardial infarction | 42 | 12.0 |
| Diabetes mellitus | 34 | 9.7 |
| Stroke | 26 | 7.4 |
| Gait characteristics mean, SD | ||
| Gait speed (cm/s) | 115.2 | 20.7 |
| Step length variability (cm) | 2.7 | 0.9 |
| Double support time variability (ms) | 20.5 | 10.6 |
| Step width variability (cm) | 2.1 | 0.7 |
| Step time variability (ms) | 21.7 | 13.0 |
| Cognitive tests, mean, SD | ||
| COWAT (number of words) | 36.2 | 13.4 |
| Stroop words (s) | 21.3 | 8.2 |
| Stroop color words (s) | 38.6 | 22.5 |
| Digit symbol coding (number correct) | 50.4 | 15.1 |
| Digit symbol Search (number correct) | 22.8 | 7.7 |
| Digit span (number correct) | 15.8 | 3.8 |
| Hopkins immediate recall (number correct) | 22.3 | 6.1 |
| Hopkins delayed recall (number correct) | 7.7 | 3.0 |
| Hopkins recognition (number correct) | 10.0 | 2.0 |
| Rey complex figure copy (number correct) | 32.0 | 5.0 |
| Rey complex figure delay (number correct) | 15.0 | 6.9 |
| Activities of daily living (ADL), mean, SD | 23.5 | 1.2 |
| Variable | ||
|---|---|---|
| Age (y), mean, SD | 71.9 | 7.1 |
| Male, n, % | 196 | 56.0 |
| Greater than high school education, n, % | 159 | 45.3 |
| BMI (kg/m2), mean, SD | 27.6 | 4.2 |
| Self-reported medical history, n, % | ||
| Hypertension | 174 | 49.7 |
| High cholesterol | 142 | 40.6 |
| Angina | 45 | 12.9 |
| Myocardial infarction | 42 | 12.0 |
| Diabetes mellitus | 34 | 9.7 |
| Stroke | 26 | 7.4 |
| Gait characteristics mean, SD | ||
| Gait speed (cm/s) | 115.2 | 20.7 |
| Step length variability (cm) | 2.7 | 0.9 |
| Double support time variability (ms) | 20.5 | 10.6 |
| Step width variability (cm) | 2.1 | 0.7 |
| Step time variability (ms) | 21.7 | 13.0 |
| Cognitive tests, mean, SD | ||
| COWAT (number of words) | 36.2 | 13.4 |
| Stroop words (s) | 21.3 | 8.2 |
| Stroop color words (s) | 38.6 | 22.5 |
| Digit symbol coding (number correct) | 50.4 | 15.1 |
| Digit symbol Search (number correct) | 22.8 | 7.7 |
| Digit span (number correct) | 15.8 | 3.8 |
| Hopkins immediate recall (number correct) | 22.3 | 6.1 |
| Hopkins delayed recall (number correct) | 7.7 | 3.0 |
| Hopkins recognition (number correct) | 10.0 | 2.0 |
| Rey complex figure copy (number correct) | 32.0 | 5.0 |
| Rey complex figure delay (number correct) | 15.0 | 6.9 |
| Activities of daily living (ADL), mean, SD | 23.5 | 1.2 |
Note: SD, standard deviation; BMI, Body Mass Index; kg, kilograms; m, meter; cm, centimeter; s, seconds; ms, milliseconds, COWAT, Controlled Word Association Test.
Characteristics of Participants (n = 350)
| Variable | ||
|---|---|---|
| Age (y), mean, SD | 71.9 | 7.1 |
| Male, n, % | 196 | 56.0 |
| Greater than high school education, n, % | 159 | 45.3 |
| BMI (kg/m2), mean, SD | 27.6 | 4.2 |
| Self-reported medical history, n, % | ||
| Hypertension | 174 | 49.7 |
| High cholesterol | 142 | 40.6 |
| Angina | 45 | 12.9 |
| Myocardial infarction | 42 | 12.0 |
| Diabetes mellitus | 34 | 9.7 |
| Stroke | 26 | 7.4 |
| Gait characteristics mean, SD | ||
| Gait speed (cm/s) | 115.2 | 20.7 |
| Step length variability (cm) | 2.7 | 0.9 |
| Double support time variability (ms) | 20.5 | 10.6 |
| Step width variability (cm) | 2.1 | 0.7 |
| Step time variability (ms) | 21.7 | 13.0 |
| Cognitive tests, mean, SD | ||
| COWAT (number of words) | 36.2 | 13.4 |
| Stroop words (s) | 21.3 | 8.2 |
| Stroop color words (s) | 38.6 | 22.5 |
| Digit symbol coding (number correct) | 50.4 | 15.1 |
| Digit symbol Search (number correct) | 22.8 | 7.7 |
| Digit span (number correct) | 15.8 | 3.8 |
| Hopkins immediate recall (number correct) | 22.3 | 6.1 |
| Hopkins delayed recall (number correct) | 7.7 | 3.0 |
| Hopkins recognition (number correct) | 10.0 | 2.0 |
| Rey complex figure copy (number correct) | 32.0 | 5.0 |
| Rey complex figure delay (number correct) | 15.0 | 6.9 |
| Activities of daily living (ADL), mean, SD | 23.5 | 1.2 |
| Variable | ||
|---|---|---|
| Age (y), mean, SD | 71.9 | 7.1 |
| Male, n, % | 196 | 56.0 |
| Greater than high school education, n, % | 159 | 45.3 |
| BMI (kg/m2), mean, SD | 27.6 | 4.2 |
| Self-reported medical history, n, % | ||
| Hypertension | 174 | 49.7 |
| High cholesterol | 142 | 40.6 |
| Angina | 45 | 12.9 |
| Myocardial infarction | 42 | 12.0 |
| Diabetes mellitus | 34 | 9.7 |
| Stroke | 26 | 7.4 |
| Gait characteristics mean, SD | ||
| Gait speed (cm/s) | 115.2 | 20.7 |
| Step length variability (cm) | 2.7 | 0.9 |
| Double support time variability (ms) | 20.5 | 10.6 |
| Step width variability (cm) | 2.1 | 0.7 |
| Step time variability (ms) | 21.7 | 13.0 |
| Cognitive tests, mean, SD | ||
| COWAT (number of words) | 36.2 | 13.4 |
| Stroop words (s) | 21.3 | 8.2 |
| Stroop color words (s) | 38.6 | 22.5 |
| Digit symbol coding (number correct) | 50.4 | 15.1 |
| Digit symbol Search (number correct) | 22.8 | 7.7 |
| Digit span (number correct) | 15.8 | 3.8 |
| Hopkins immediate recall (number correct) | 22.3 | 6.1 |
| Hopkins delayed recall (number correct) | 7.7 | 3.0 |
| Hopkins recognition (number correct) | 10.0 | 2.0 |
| Rey complex figure copy (number correct) | 32.0 | 5.0 |
| Rey complex figure delay (number correct) | 15.0 | 6.9 |
| Activities of daily living (ADL), mean, SD | 23.5 | 1.2 |
Note: SD, standard deviation; BMI, Body Mass Index; kg, kilograms; m, meter; cm, centimeter; s, seconds; ms, milliseconds, COWAT, Controlled Word Association Test.
Regional Mean Cortical Thickness and Gait Variability
The specific cortical regions associated with each gait variability measure are presented in detail in Supplementary Table S1. Smaller cortical thickness in widespread brain regions including frontal (bilateral precentral [primary motor], bilateral superior frontal [supplementary motor], left inferior frontal, right middle frontal [prefrontal]), parietal (bilateral inferior parietal, left precuneus), temporal (left superior, middle, inferior temporal), and occipital areas (bilateral lateral occipital) was associated with greater variability in step width and step time (Figures 1 and 2). In addition, parietal (bilateral postcentral [primary somatosensory], bilateral supramarginal, right superior parietal, right precuneus), left insula, temporal (right fusiform), and occipital (right lingual) regions were associated with step width variability, and additional frontal (left frontal pole, right lateral orbitofrontal) regions were associated with step time variability. Smaller cortical thickness in only frontal (left inferior frontal) and temporal (left superior, middle temporal) regions was associated with greater DST variability (Figure 3). smaller thickness in only the occipital region (right cuneus) was associated with greater step length variability (Figure 4).
Brain regions associated with step width variability. Top panels show the areas identified by the mean cortical thickness and the bottom panels show the areas identified by cortical thickness ratio; color scale corresponds to log of one-sided Bayes factors; regions of cool colors indicate smaller thickness or ratio associated with greater variability; regions of warmer colors indicate greater thickness or ratio associated with greater variability. Only the regions with credibility intervals that do not cross zero are shown, A-P; Anterior-Posterior, RH; Right Hemisphere, LH; Left Hemisphere.
Brain regions associated with step time variability. Top panels show the areas identified by the mean cortical thickness and the bottom panels show the areas identified by cortical thickness ratio; color scale corresponds to log of one-sided Bayes factors; regions of cool colors indicate smaller thickness or ratio associated with greater variability; regions of warmer colors indicate greater thickness or ratio associated with greater variability. Only the regions with credibility intervals that do not cross zero are shown, A-P; Anterior-Posterior, RH; Right Hemisphere, LH; Left Hemisphere.
Brain regions associated with double support time variability. Top panels show the areas identified by the mean cortical thickness and the bottom panels show the areas identified by cortical thickness ratio; color scale corresponds to log of one-sided Bayes factors; regions of cool colors indicate smaller thickness or ratio associated with greater variability; regions of warmer colors indicate greater thickness or ratio associated with greater variability. Only the regions with credibility intervals that do not cross zero are shown, A-P; Anterior-Posterior, RH; Right Hemisphere, LH; Left Hemisphere.
Brain regions associated with step length variability. Top panels show the areas identified by the mean cortical thickness and the bottom panels show the areas identified by cortical thickness ratio; color scale corresponds to log of one-sided Bayes factors; regions of cool colors indicate smaller thickness or ratio associated with greater variability; regions of warmer colors indicate greater thickness or ratio associated with greater variability. Only the regions with credibility intervals that do not cross zero are shown, A-P; Anterior-Posterior, RH; Right Hemisphere, LH; Left Hemisphere.
Greater thickness in the right anterior cingulate was associated with greater variability in all measures, and greater thickness in the left anterior cingulate region was associated with greater DST and step length variability (Figures 1–4).
Regional Cortical Thickness Ratio and Gait Variability
The cortical thickness ratio regions associated with each gait variability measure are presented in detail in Supplementary Table S2. Smaller cortical thickness ratio in frontal (bilateral precentral, left middle frontal), temporal (left superior), and parietal (right inferior) regions was associated with greater step width variability (Figure 1) while a smaller ratio in the frontal (bilateral precentral, left frontal pole, left inferior frontal) and temporal (left superior) regions was associated with greater step time variability (Figure 2). A smaller thickness ratio in frontal (bilateral precentral, left inferior frontal) and temporal (left superior and middle temporal) regions was associated with greater DST variability (Figure 3). A smaller ratio in frontal (right middle frontal), temporal (left middle temporal), and parietal (left postcentral, right precuneus) and occipital (right cuneus) regions was associated with greater variability step length (Figure 4).
In contrast, a higher cortical thickness ratio in the cingulate (right anterior, right posterior, isthumus of cingulate), frontal (right inferior frontal, right medial orbtiofrontal), and temporal (right middle temporal) regions was associated with greater variability in most measures (Figures 1–4). Additionally, a higher thickness ratio in frontal (left medial orbitofrontal, right lateral orbitofrontal), occipital (left cuneus) and the right insula was associated with greater step width variability, and a greater ratio in frontal (left inferior frontal), temporal (right parahippocampal gyrus) and left posterior cingulate regions was associated with greater step time variability.
Regional Thickness Measures and Gait Speed
Smaller mean cortical thickness in frontal (bilateral precentral, bilateral superior frontal, bilateral middle frontal, left inferior frontal, right lateral orbitofrontal), temporal (bilateral superior, left middle temporal), and parietal (right superior parietal, right precuneus) regions was associated with slower gait speed. A smaller cortical thickness ratio in similar, but fewer, regions were also associated with slower gait speed. In contrast, a higher thickness ratio in the cingulate (bilateral anterior, isthmus, and left posterior cingulate), frontal (left inferior frontal), and temporal (left fusiform, right parahippocampal) regions was associated with slower gait speed (Figure 5).
Brain regions associated with gait speed. Top panel shows the areas identified by the mean cortical thickness and the bottom panel shows the areas identified by cortical thickness ratio; color scale corresponds to log of one-sided Bayes factors; regions of cooler colors indicate smaller thickness or ratio associated with slower speed; regions of warmer colors indicate greater thickness or ratio associated with slower speed. Only the regions with credibility intervals that do not cross zero are shown. A-P; Anterior-Posterior, RH; Right Hemisphere, LH; Left Hemisphere.
Supplementary Table S3 (regional cortical thickness) and Supplementary Table S4 (cortical thickness ratio) show the cortical regions associated with each gait measure only in participants whose scores were >1.5 SD of age, sex, and education appropriate norms in all cognitive domains (n = 222).
Discussion
This study is the first to the best of our knowledge to examine associations between cortical thickness and gait variability in community-dwelling older people. Smaller overall cortical thickness (the mean of all regions) was only associated with greater variability in two measures—step width and step time. In contrast, regional cortical thickness in distinct areas was associated with all gait variability measures, highlighting the importance of examining individual regions and different gait measures. The thickness ratio identified some similar but also different regions, suggesting that smaller thickness in brain regions relative to the global mean might also be important for consistency in gait. Together, our findings add to the knowledge of cortical control of gait variability in older age.
Associations Between Areas of Smaller Cortical Thickness and Greater Gait Variability
Smaller thickness overall and in widespread regions was associated with step width and step time variability. These included areas important for motor planning (bilateral prefrontal and bilateral supplementary motor) and execution (bilateral primary motor), executive function (bilateral prefrontal) (32,33), perception and integration of visuospatial inputs (bilateral inferior parietal, left precuneus) (34,35), vestibular processing (left superior temporal) (36), auditory and movement perception (left middle temporal) (37) and visual object recognition (bilateral lateral occipital, left inferior temporal) (37). This is consistent with the knowledge that a wide range of cognitive (particularly executive function) (7) and sensorimotor functions that require central processing (ie, postural sway, reaction time, proprioception) (8) are important for the control of gait. However, in prior work, poorer performance in cognition was not associated with step width variability, while only poorer executive function was associated with greater step time variability (7). The widespread brain regions associated with these measures suggest that cortical thinning may impact on gait variability at an earlier stage than can be detected by neuropsychological tests.
For step width variability, additional regions included areas important for motor planning and updating (bilateral and a greater number of prefrontal regions), processing of visual (right fusiform, right lingual gyrus, right superior parietal, bilateral supramarginal) (36,37), vestibular (left insula) (36), and somatosensory (bilateral primary somatosensory, supramarginal) information (38). These additional regions may suggest that balance control in the coronal plane (reflected by step width variability) requires higher levels of attention and sensory integration. In line with our results, lower parietal volume was associated with greater stride time variability in a prior study (12). However, prior studies of stride width variability have only examined hippocampal volume (not included in our study) (13), and therefore cannot be compared with our results. Interestingly, the regions associated with these variability measures in our study are similar, but greater in number (particularly in frontal areas), when compared to regions associated with absolute step width and step time in prior studies (20,39). This suggests that frontal areas might be particularly important in maintaining consistency in timing of steps and balance control in the coronal plane. For both step width and step time variability, most regions associated with mean thickness disappeared when replaced with the cortical thickness ratio. Taken together, our findings suggest that smaller thickness in widespread brain regions is important for maintaining the consistency in step width and step time, rather than specific areas of smaller regional thickness in relation to overall mean thickness.
In contrast to step width and step time variability, fewer regional areas were associated with DST and step length variability and there was no association with overall mean thickness. The thickness ratio identified more regions than the mean thickness measure, including areas important for motor planning (prefrontal), auditory processing (left middle temporal), execution of movement (bilateral primary motor), vestibular function (left superior temporal) (for DST variability), vision (right precuneus, right cuneus) and somatosensory function (left primary somatosensory) (for step length variability). Interestingly many of these areas have been found to atrophy at a greater rate than other areas of the brain with advancing age (22,40). Our findings suggest that smaller thickness in these regions relative to the rest of the cortex is important for step-to-step consistency in DST and step length.
In prior studies from our cohort, DST and step length variability were associated with a wide range of cognitive and sensorimotor functions (7,8). Therefore, it is somewhat surprising that the cerebral regions associated with these measures were largely limited to frontal and temporal regions. Considering previous associations between smaller subcortical gray matter volumes and step length (16,24,41), and prior findings of an association between DST variability and memory decline (42), it is possible that subcortical regions (ie, the hippocampus, basal ganglia, the cerebellum) might play a role in the control of DST and step length variability. However, the surface-based stream of FreeSurfer only includes cortical regions; therefore, we did not examine subcortical regions. Furthermore, compared to variability in DST, smaller volume in widespread frontal and sensory regions were associated with absolute DST in a prior study (39). Similarly, smaller volume and thickness in a greater number of cortical regions was associated with absolute step length, than step length variability in our study (20,24,39). However, results are difficult to compare as these prior studies examined cortical thickness only in older people with cerebral small vessel disease (20) or examined gray matter volume (24,39), rather than thickness.
Associations Between Areas of Greater Cortical Thickness and Greater Gait Variability
Unexpectedly greater mean thickness and a higher thickness ratio in the anterior cingulate was associated with greater variability in all measures. In addition, a greater thickness ratio in right prefrontal and right middle temporal regions was associated with greater variability in most measures. We are unsure of the mechanism underlying these findings. A possible reason may be that greater (or relatively preserved) thickness is a neuroplastic adaptation to maintain gait consistency, compensating for areas of smaller thickness. Functions of the anterior cingulate include allocation of attentional resources, planning, decision making, and error monitoring and correction (43,44), which are important for gait control (7). Imaging studies have also reported greater activation in the anterior cingulate during dual-task walking (45). This is only speculation but supported by inverse associations between the thickness ratio in the cingulate with other regions in our study (ie, right caudal anterior cingulate correlated inversely with right precentral [r = −.25] and right superior parietal [r = −.26] regions). However, compensation may have only been helpful in mitigating impairment to a point, as greater thickness was not associated with better performance.
Regional Cortical Thickness and Gait Speed
We also examined associations between cortical thickness and gait speed to enable comparison with gait variability and prior studies of gait speed. Overall mean thickness was not associated with gait speed, which is in line with cross-sectional findings from a recent study (46). Similar to prior a study, smaller mean thickness in frontal, temporal, and parietal regions was associated with slower gait speed (20). However, compared to our results, the regions in the prior study were more dispersed in the parietal and occipital areas. This might be due to the prior study sample, including only older people with cerebral small vessel disease, as the strength of some associations were attenuated after adjustments for subcortical vascular lesions. Also, in contrast to their findings, we found a higher cortical thickness ratio in the cingulate was associated with slow gait speed.
Compared to associations with variability in step time and step width, a smaller number of sensory regions were associated with gait speed, suggesting these areas are important for consistency in timing and balance control to a greater extent than gait speed. In contrast, compared to variability in DST and step length, a greater number of frontal regions were associated with gait speed. This is consistent with knowledge that poorer attention and executive function are associated with slower gait speed (7).
Strength and Limitations
The examination of associations between regional cortical thickness and gait variability is novel. Our analysis has the advantage of examining multiple regions, in contrast to others that have examined gray matter volume only in a few pre-specified regions of interest. Cortical thickness is a sensitive marker of morphological aging that is not associated with total intracranial volume (18). Participants were a randomly selected sample of older people from the electoral roll, and therefore the results may be more generalizable to a wider community of older people than other studies of participants from memory clinics (13,14). The use of the cortical thickness ratio allowed us to determine whether regions with smaller thickness relative to the rest of the cortex were important for gait consistency. There are, however, limitations to our study. Although the surface-based stream of FreeSurfer5.3 provides a detailed list of brain regions, it does not include the hippocampus (lies in the cortical infolding) or subcortical areas. The design of this study limits our inferences to cross-sectional associations; thus, we cannot provide insights as to whether the decline in thickness in these regions are contributing to gait variability. For example, a recent study found that smaller global mean thickness was associated with gait slowing over time (46). It is possible that our sample may have included older people with mild cognitive impairment, which may have modified associations between cortical thickness and gait. However, we did not have a formal diagnosis for mild cognitive impairment. Instead we present the results for those with no cognitive impairment on neuropsychological testing. Generally, compared to the whole sample, fewer regions with smaller cortical thickness and more regions with greater cortical thickness (including the anterior cingulate) were associated with greater gait variability. However, comparison between results should be viewed with caution as the number of participants differed. Also, a limitation in this study is that we measured gait variability over a mean of 27.9 steps, which may have affected the reliability of gait variability measures. However, despite this, we were still able to find a number of associations.
Conclusion
This study provides the first evidence for associations between cortical thickness and intra-individual gait variability in older people. Smaller overall cortical thickness was only associated with greater variability in step width and step time. In contrast, smaller regional thickness in areas important for sensorimotor and cognitive functions was associated with greater variability in all measures. However, brain regions associated with each gait variability measure were specific. Our findings highlight the importance of examining individual brain regions to understand the cortical control of distinct gait variability measures.
Funding
This work was supported by National Health and Medical Research Council (NHMRC; grant number 403000BH and 491109), Physiotherapy Research Foundation (grant number BH036/05), Perpetual Trustees, Brain Foundation, Royal Hobart Hospital Research Foundation (grant number 341M), Australia and New Zealand Charitable Trust, Masonic Centenary Medical Research Foundation, VS is funded by a NHMRC Practitioner Fellowship (APP1137837), MC is funded by a NHMRC Boosting Dementia Research Leadership Fellowship (1135761).
Acknowledgments
Study participants: Tasmanian Study of Cognition and Gait research staff and volunteers. The computational resources provided by the Australian Government through MASSIVE under the National Computational Merit Allocation Scheme.
Author Contributions
Author contributions included conception and study design (V.K.S., M.L.C.), data collection or acquisition (V.K.S., M.L.C.), statistical analysis (O.J., R.B.), interpretation of results (O.J., R.B., M.L.C.), drafting the manuscript work or revising it critically for important intellectual content (O.J., R.B., M.L.C., M.B., H.M.B., V.K.S.) and approval of final version to be published and agreement to be accountable for the integrity and accuracy of all aspects of the work (All authors).
Conflict of Interest
None reported.
