White Matter Hyperintensities Are Associated with Slower Gait Speed in Older Adults without Dementia

Abstract

Introduction

Slow gait speed is associated with poor health outcomes in aging, but the relationship between cerebral small vessel disease (CSVD) pathologies and gait speed in aging is not well understood. We investigated the relationships between CSVD imaging markers and gait speed during simple (normal pace walking [NPW]) and complex (walking while talking [WWT]) as both measures are associated with shared health outcomes such as falls, frailty, disability, mortality, and dementia.

Methods

A total of 113 Ashkenazi Jewish adults over 65 (M age = 78.6 ± 6.3 years, 45.8% women) and without dementia were examined. Established rating systems were used to quantify white matter hyperintensities (WMHs) and lacunes of presumed vascular origin from fluid-attenuated inversion recovery (FLAIR) images. Linear regression models adjusted for age, sex, global health, and total intracranial volume were used to examine associations between CSVD markers and gait speed during NPW and WWT. Student t tests were used to contrast gait speed in those with “confluent-diffuse” WMH and those with “mild or no” WMH.

Results

The number of WMH in the basal ganglia (β = −3.274 cm/s p = 0.047) and temporal lobes (β = −3.113 cm/s p = 0.048) were associated with slower NPW speed in adjusted models. Participants with higher CSVD burden (confluent-diffuse pattern) in the frontal lobe (94.65 cm/s vs. 105.21 cm/s, p = 0.018) and globally (98.98 cm/s vs. 107.24 cm/s, p = 0.028) also had lower NPW speed. WMHs were not associated with WWT speeds. Lacunes were not associated with NPW or WWT speed.

Conclusion

Adjusted models found higher CSVD burden as measured by the presence of WMH in the basal ganglia and temporal lobes were associated with slower normal pace gait speed in older adults, but not with complex walking speeds. Participants with confluent-diffuse WMHs in the frontal lobes were found to have slower average normal gait speed. Further studies are needed to establish the temporality of WMH and gait speed decline as well as mechanistic links between the two.

Key Points

• Basal ganglia and temporal lobe WMHs were associated with slower gait speed in older adults.

• Participants with confluent-diffuse WMH in the frontal lobes had slower average gait speed.

Introduction

Measurable parameters of gait function, such as gait speed, arm swing, and stride length and width, are correlated with overall physical function and health in the aging population [1]. Gait speed, in particular, is as accurate a predictor of survival and life expectancy in older adults as the use of mobility aids and self-reported function [2]. Slower gait speed has been associated with age, decreased literacy, difficulty with IADLs, and cardiovascular disease [3]. Slow gait speed has also been found to predict cognitive decline in older adults, being detectable years before diagnosing cognitive impairment or dementia [4–6]. For instance, one study found that individuals who develop mild cognitive impairment (MCI), a pre-dementia syndrome, show earlier and greater changes in the rate of gait speed decline than those who do not develop MCI. This decline in gait speed was observed years before MCI was detected [7]. It has also been shown that individuals with slower gait speed at baseline have an increased risk of developing both Alzheimer’s and all-cause dementia later in life [8]. These findings indicate that gait speed can be a powerful clinical tool for predicting adverse physical and cognitive outcome risk in older, cognitively intact individuals [9].

Cerebral small vessel disease (CSVD) describes pathology affecting the perforating arterioles and capillaries supplying the brain’s deep structures including subcortical white matter and deep gray matter. CSVD lesions are often clinically silent in aging brains but can be seen on brain imaging as white matter hyperintensities (WMHs), lacunes, microbleeds, enlarged perivascular spaces, and microinfarcts. While the pathogenesis of these lesions is not well understood, WMHs are thought to represent areas of demyelination, axon loss, and gliosis, while lacunes are fluid-filled cavities replacing damaged tissue, and microbleeds are thought to be areas of fixed hemorrhages [10]. Studies have found associations between the presence and severity of CSVD and cognitive impairment [11], brain atrophy [12], as well as gait disturbances [13]. For instance, periventricular and frontal lobe WMHs are associated with a history of falls in an elderly, non-disabled population [14]. Investigation into whether CSVD mediates gait speed has produced inconsistent findings to date. For example, global WMH score, i.e., total WMH burden across all brain regions, but not cerebral microbleeds or lacunes, was associated with slower gait speed in healthy, older adults [15]. Another study found that the presence of cerebral microbleeds and/or lacunes, but not overall WMH volume, was associated with slower gait speed [16]. A recent cross-sectional study also showed a positive association between dual-task cost (DTC) (relative difference between normal pace walking (NPW) speed and walking speed while performing a cognitive task) and WMH burden in two subcortical regions: basal ganglia and thalamus. Of note, all participants included in this study had a history of stroke [17].

In this study of older adults without dementia, we aimed to explore the association between CSVD burden and gait speed. To do so, we took a semiquantitative and regional approach to characterize CSVD burden and examine the effects of both severity and location of these markers on gait. For WMH, we also looked at the association between more severe WMH burden (defined as confluent or diffuse lesions) versus mild or no WMH in different brain regions given that overall greater WMH volume is associated with gait disturbances [18]. We focus here on ischemic CSVD imaging markers (WMH and lacunes) rather than hemorrhagic (microbleeds) due to the availability of imaging parameters to detect the former. We examined both NPW speed and walking while talking (WWT) speed – a cognitively more demanding mobility condition that involves walking while simultaneously reciting alternate letters of the alphabet [19]. These two walking conditions were chosen given that longitudinal studies have shown lower speeds in both tasks are independently associated with worse cognitive and functional outcomes in older adults [20, 21]. We hypothesized that greater WMH and lacune burden in frontal regions would be associated with slower gait speed in both walking conditions and that this effect would be greater with WWT conditions given the higher cognitive load this task employs. We also hypothesized that participants with confluent-diffuse WMHs in the frontal lobes would have slower average gait speed given that histopathologic studies have suggested confluent-diffuse lesions represent areas of more severe ischemic tissue damage in relation to punctate or focal WMHs, lesions that are much more frequent in prevalence [22]. Elucidating the association between imaging markers of CSVD and gait may indicate those with slow gait are at higher risk of having underlying CSVD. Therefore, prevention of CSVD may in turn improve gait and overall physical and cognitive function in aging adults. In addition, measuring CSVD burden by brain region may help to better understand the brain pathology underlying predisposition to gait speed decline in aging.

Methods

Cohort

The LonGenity study is a longitudinal cohort of community-dwelling Ashkenazi Jewish older adults that examines genetic and biological mechanisms that protect from age-related diseases and promote successful aging [23]. LonGenity study participants are defined as either offspring of parents with exceptional longevity (having at least 1 parent lived to age 95 or older) or offspring of parents with usual survival. Individuals with a diagnosis of dementia (>8 on the Blessed Mental Status Examination and >2 on the AD8-item Informant Questionnaire) at the time of imaging, severe visual or hearing impairment, and siblings already enrolled in the study were excluded. A subset of 113 LonGenity study participants who had neuroimaging (MRI) done as part of a sub-study were included in this investigation. Ethical clearance was obtained from the Committee on Clinical Investigations of the Albert Einstein College of Medicine. Written informed consent was obtained from all participants.

MRI Acquisition and Processing

A Philips 3T multinuclear MRI/MRS system was used to acquire 3D-FLAIR images (N = 89; TR/TI = 4,300/1,650 ms, TE = 325 ms, 240 × 238 acquisition matrix, and 0.46 mm voxel size) or 2D-FLAIR images (N = 24; TR/TI = 11,000/2,800 ms, TE = 125 ms, 250 × 250 acquisition matrix, and 0.40 mm voxel size). T1-weighted images were also acquired using the following parameters: 240 × 220 × 240 mm FOV; 1.0 mm isotropic resolution; TR/TE = 9.8/4.6 ms; SENSE factor = 2.5. A total of 113 FLAIR/T1 images had corresponding NPW velocity measurements and were rated for WMH and lacune burden.

CSVD markers investigated in this study included WMH and lacunes due to these markers’ associations with gait function [18]. Cerebral microbleeds were not examined in this study given the lack of T2*-weighted and susceptibility-weighted imaging. Given their high prevalence in older populations, less is known about the role of enlarged perivascular spaces in gait function and were not investigated in this study [24].

CSVD Quantification

Lacunes of presumed vascular origin were identified as focal, round, or ovoid, fluid-filled lesions 3–15 mm in diameter with a hyperintense rim on FLAIR images [24]. Lesions had to be hypointense on T1 and FLAIR images to be classified as lacunes. Brain regions measured for total number of lacunes included frontal, parietal, temporal, occipital lobes, infratentorial white matter (brainstem and cerebellum), and deep brain regions (periventricular, basal ganglia, thalamus, internal and external capsule, and corpus callosum). Total numbers of lacunes were summed for infratentorial and deep regions as well as a global lacune score.

WMHs were quantified using the Age-Related White Matter Changes scale (ARWMC) in five regions: frontal, parietal-occipital (parietal and occipital lobes), temporal, infratentorial, and basal ganglia [25]. The ARWMC scale has been validated as a simplified score that correlates well with computer-aided volumetric measures [26]. Scores in all ten regions (5 for each side) were summed to obtain a global WMH burden score (range 0–30). Briefly, focal WMHs were given a rating of 1, confluent WMHs were given a rating of 2, and diffuse WMHs were given a rating of 3. For the basal ganglia, focal WMHs were given a rating of 1, while a rating of 2 was given if multifocal WMHs were detected, and confluent WMHs were given a rating of 3.

Ratings for both lacunes and WMH’s were manually performed by author JV after being trained in CSVD quantification by a board-certified neurologist over a period of 2 months. Interrater reliability was calculated using a random subset of images from a cohort not used in this study with substantial agreement between both raters (n = 30, linearly weighted κ = 0.80, κ = 0.66 for WMH and lacunes, respectively). All ratings were performed blinded to participant demographics and other measures.

Gait Assessment

Gait speed (cm/s) during usual pace walking and WWT (walking while reciting alternate letters of the alphabet [27]) was assessed using validated protocols on an 8.5-meter-long computerized GAITRite walkway (following a practice walk). To allow for initial acceleration and terminal deceleration, the walk started 1.2 m before and ended 1.2 m after the walkway. High interrater and test-retest reliability has been reported for both walking protocols in our center [28]. Participants wore comfortable footwear and walked in a quiet and well-lit room. No assistive devices were used during the walk trials. Finally, DTC was defined as the percentage difference between NPW and WWT velocity and calculated as follows:

$$\text{DTC}=\frac{\text{NPW}-\text{WWT}}{\text{NPW}}\times 100$$

Cognitive Measures

Global cognitive score was calculated as a composite score of 11 neuropsychological tests with higher scores indicating better overall cognition spanning cognitive domains including executive function, working memory, processing speed, language, visuospatial function, visual memory, and verbal episodic memory. Trail-Making Test A time was used as a measure of processing speed, while Trail-Making Test B-A (TMT interference) was used a surrogate for executive function as previously described [29, 30].

Covariates

Demographic data like age and sex were recorded at initial screening. The summed value of the presence or absence of 9 medical conditions (hypertension, diabetes, chronic heart failure, arthritis, depression, stroke, chronic obstructive pulmonary disease, myocardial infarction, angina) was calculated to create a global health score. Estimated total intracranial volume as well as regional cortical thickness and volume measurements were obtained from structural MRIs using FreeSurfer 6.0 software [31, 32].

Statistical Analysis

All analyses were performed using the statistical software Stata version 17.0. Linear regression models adjusted for covariates (age, sex, global health score, parental longevity, and estimated total intracranial volume) were used to investigate the association between measures of CSVD (lacunes and WMH) and gait speed with NPW and WWT. A priori α level was set at p = 0.05. Two-sample Student’s t test was performed to compare means of velocities between confluent-diffuse WMH versus focal or no WMH. Participants were separated into two groups for each region: focal or no WMH (defined as WMH rating <2) or confluent-diffuse WMH (defined as WMH rating >2). Multiple comparison correction was used to decrease false discovery rate for the two-sample Student’s t test (Bonferroni method, corrected p value of 0.05/2 = 0.025).

Results

Our cohort included a total of 113 participants who had FLAIR imaging and CSVD ratings. Of these 113 participants, all had NPW speed measured, and 99 had WWT speed measured at the same visit at which MRI was done. Both outcomes were normally distributed between both groups in our cohort (Shapiro-Wilk test for normality, p = 0.36, p = 0.80 for NPW and WWT speed, respectively). Demographic data are summarized in Table 1.

Table 1.

Participant demographics grouped by walking task

Characteristics	NPW (n = 113)	WWT (n = 99)
Age, mean (SD), years	78.6 (6.3)	78.6 (6.1)
Sex female, %	45.8	53.5
Diabetes, %	8.9	7.1
Hypertension, %	28.6	27.6
Global health score, mean (SD)	0.39 (0.63)	0.36 (0.63)
Estimated total intracranial volume, mean, cm3	1,416	1,425
Walking speed, mean (SD), cm/s	102.8 (19.9)	75.2 (21.9)
Overall WMH, mean (IQR)	5.60 (2–8)	5.51 (2–8)
Overall lacunes, mean (IQR)	0.34 (0–1)	0.36 (0–1)
Overall cognitive score, mean (SD)	2 (6.7)	2.3 (6.4)
Trail-making time A, mean (SD), s	40.5 (16.2)	39.3 (13.3)
Trail-making time interference, mean (SD), s	48.9 (36.5)	47.4 (34.9)

SD, standard deviation; IQR, interquartile range.

WMH burden in both the bilateral basal ganglia and the temporal lobes were associated with slower NPW speed in the fully adjusted models (β = −3.274 cm/s, p = 0.047, β = −3.113 cm/s, p = 0.048, respectively) (Table 2). Figure 1 shows representative images of bilateral temporal and basal ganglia WMHs. Higher frontal lobe WMH burden was associated with slower NPW speed in the unadjusted model (β = −3.316 cm/s, p = 0.017) but not in the fully adjusted model (β = −1.521 cm/s, p = 0.262). WMH ratings were not found to be significantly associated with WWT speed. Lacunes were not significantly associated with either NPW or WWT speed in any region.

Table 2.

Linear regression analysis between number of WMH in select brain regions and NPW

Location of white matter/subcortical hyperintensities	Coefficient, cm/s	95% CI	p value
Frontal	−1.52	(−4.20, 1.51)	0.262
Parieto-occipital	−1.10	(−3.55, 1.36)	0.380
Temporal	−3.11	(−6.20, −0.025)	0.048
Basal ganglia	−3.27	(−6.50, −0.047)	0.047
Infratentorial	−1.09	(0.5.47, 3.29)	0.623
Overall	−0.77	(−1.57, 0.059)	0.069

Models were adjusted for age, sex, recruitment status, global health score, and estimated intracranial volume. Calculated coefficients are expressed in cm/s.

Bolded values indicate a calculated p value <0.05.

Fig. 1.

Representative axial FLAIR images showing CSVD lesions. WMH indicated by red arrows. a Bilateral focal WMH in the basal ganglia. b Bilateral focal WMH in the subcortical temporal lobes. c Bilateral confluent frontal lobe WMH. d Bilateral focal frontal lobe WMH.

We then examined whether average gait speed differed as a function of WMH severity in all brain regions. Table 3 shows participants with confluent-diffuse WMHs in the frontal lobes had slower average NPW speed when compared to those with focal or no WMH (94.65 vs. 105.21 cm/s, p = 0.018) but not with WWT speed. We also examined if having a higher “global” WMH burden, i.e., having an overall WMH rating >4 in all brain regions, was associated with NPW and WWT speed compared to the rest of the sample; however, this difference did not reach statistical significance after correcting for multiple comparisons (98.98 vs. 107.24, p = 0.028). Figure 1 shows representative images of focal and confluent-diffuse lesions on FLAIR imaging in our cohort.

Table 3.

Comparisons of average gait speed between groups having confluent-diffuse WMH versus focal or no WMH

Region	NPW				WWT
	presence of confluent-diffuse WMHa	participants, n	average gait speed, cm/s	p value	presence of confluent-diffuse WMHa	participants, n	average gait speed, cm/s	p value
Frontal	Yes	26	94.65	0.018	Yes	23	74.34	0.839
	No	87	105.21		No	76	75.41
Parieto-occipital	Yes	25	100.77	0.571	Yes	22	77.50	0.573
	No	88	103.36		No	77	74.49
Temporal	Yes	4	102.78	0.981	Yes	4	70.08	0.639
	No	109	103.03		No	95	75.37
Basal gangliab	Yes	6	101.27	0.849	Yes	5	81.44	0.515
	No	107	102.87		No	94	74.83
Infratentorial	Yes	32	99.40	0.259	Yes	25	77.02	0.626
	No	81	104.12		No	74	74.53
Overallc	Yes	61	98.98	0.028	Yes	52	76.15	0.641
	No	52	107.24		No	47	74.07

^aParticipants with confluent-diffuse WMH (rating >2) in each region were grouped in the “yes” category.

^bCategorized as “yes” for presence of multifocal hyperintensities (rating >2).

^cCategorized as “yes” if global WMH rating >4.

Bolded values indicate p value <0.05.

We performed sensitivity analyses examining whether cortical thickness or volume of the brain regions of interest may confound the associations between WMHs and NPW velocities found in our models. Online supplementary Table 1 (for all online suppl. material, see https://doi.org/10.1159/000538944) shows that including regional cortical thickness or volume in our models did not modify the significant associations found in our original models in Table 2, suggesting that the effects of WMHs on velocity are independent of these measurements. Online supplementary Table 2 looked at DTC as an outcome variable instead of NPW or WWT velocity, given DTC has also been shown to be an effective predictor of motor and cognitive disabilities in aging populations [20]; however, significant associations were not found. Finally, online supplementary Table 3 shows models examining the associations between overall and executive cognitive performance and WMHs and found that lesions in the frontal, temporal, and infratentorial regions were associated with worse overall cognition but not with executive function (β = −1.13, p = 0.026, β = −1.31, p = 0.042, β = −2.49, p = 0.005).

Discussion

The current study shows that CSVD lesions in the basal ganglia and temporal white matter were associated with slower gait speed in older adults without dementia, even after adjusting for age, sex, parental longevity, and intracranial volume. Average NPW speed was slower among participants with higher WMH burden compared to participants with mild or no WMH burden in frontal lobe regions. We did not find associations between markers of CSVD and WWT speed in this study or associations between lacunes and gait speed in either walking condition.

Gait is a complex function requiring both motor and cognitive integration across various brain regions [33]. There is evidence that suggests that CSVD pathology may have a causative, deleterious effect on connectivity between brain regions involved in gait from a functional level. For instance, an fMRI study found aberrant connections between the supplementary motor area of the frontal lobe and various regions of the temporal lobe in patients with CSVD and gait disorders [34]. The basal ganglia are a set of deep gray matter nuclei critical for regulating and planning motor movements. As a deep brain structure with high metabolic demand, the basal ganglia are particularly susceptible to ischemia and dysfunction due to CSVD. One study has found that decreased basal ganglia resting-state connectivity is associated with slower gait speed. This decrease in resting-state connectivity was also found to be associated with global CSVD burden [35]. It is therefore plausible that CSVD lesions in the basal ganglia may decrease functional connectivity in this structure, leading to slower gait. Post hoc analyses in our study did not show an association between WMHs and cortical thickness of regions of interest. We did, however, find higher WMH ratings in frontal, temporal, and basal ganglia regions were associated with lower overall cognitive scores. Statistically significant associations were not seen, however, for measures of executive function and processing speed. This suggests that effects of WMHs on gait speed may be mediated through cognitive impairment, an observation that has been reported in the literature [36], but through other cognitive domains such as visuospatial function or memory. A causative link requires further exploration in future studies. When global cognition was included as a covariate in our models, similar results were seen albeit without statistical significance, an unsurprising finding given that gait and cognition are interrelated measures in aging populations [37, 38]. Our finding of higher global CSVD burden being associated with slower gait speed is in line with previous studies [39].

Interestingly, our study did not show an association between either CSVD marker measured and WWT speed. This is surprising given that dual-task walking requires a higher cognitive load to execute. One study comparing participants with CADASIL, a genetic arteriopathy predisposing patients to widespread CSVD, did not show an association between CSVD burden and dual task cost, or the relative difference in performance between NPW and WWT. A potential explanation is that WWT speed declines at a slower rate and later in life (relative to our cohort average age) than NPW speed [40]. In addition, our analyses did not show an association between lacunes and gait speed. A previous study was able to show that individuals with lacunar infarcts in the frontal lobe were more likely to have motoric cognitive risk syndrome, a pre-dementia syndrome characterized by the presence of slow gait and subjective cognitive complaints [41]. One possible explanation for the lack of association in the present study is the relatively low number of lacunes found on imaging in our cohort, leading to type 2 statistical error.

Strengths of this study include the comprehensive neuroimaging characterization of WHM and lacunar burden in this cohort along with the integration of various biological and health-related parameters that may contribute to declines in gait speed among healthy, older adults. Limitations of this study include a smaller sample size and ethnic/racial homogeneity of the present cohort. We also did not examine other imaging markers of CSVD such as cerebral microbleeds and enlarged perivascular spaces. The cross-sectional design limits examination of temporality and causation. It will be worthwhile to explore the association between imaging markers of CSVD and gait and how these clinical markers evolve across time in relation to other measures like cognition.

Conclusion

We found that WMHs in temporal and basal ganglia regions were associated with slower NPW speed, and participants with confluent-diffuse lesions in frontal regions had slower average NPW speed. These associations were found to be independent of gray matter measures of these brain regions and may be mediated through CSVD effects on overall cognition. These findings have important clinical implications by adding to our understanding of how CSVD can affect gait by providing potential brain regions that may be involved in this previously observed association. This understanding may then inform how we manage care for older adults at risk of CSVD-related gait complications based on relatively simple clinical measurements.

Impact Statement

White matter hyperintensity burden is associated with slower gait speed in the frontal and temporal lobes and the basal ganglia in older adults. The novelty of these findings include the association of lesions in the temporal lobe and slower speed in older adults without dementia.

Statement of Ethics

This study protocol was reviewed and approved by the Institutional Review Board at Albert Einstein College of Medicine, NY, Approval No. 2007-272. Written informed consent was obtained from all participants prior to study enrollment.

Conflict of Interest Statement

The authors have no conflicts of interest to declare.

Funding Sources

This work was supported by grants from National Institute of Health (NIH) R01AG062659-01A1, NIH/National Institute on Aging (PI: Helena M. Blumen), R01AG057548-01A1 (PI: Joe Verghese), R01AG061155-01 (PI: Sofiya Milman), and R01AG057909 (PI: Nir Barzilai). The funder had no role in the design, data collection, data analysis, and reporting of this study.

Author Contributions

J.P.V. and H.M.B. directly contributed to the study concept and design, analysis and interpretation of data, and preparation of the manuscript. J.P.V., N.B., and S.M. directly contributed to acquisition of subjects and data, analysis and interpretation of data, and preparation of the manuscript.

Funding Statement

This work was supported by grants from National Institute of Health (NIH) R01AG062659-01A1, NIH/National Institute on Aging (PI: Helena M. Blumen), R01AG057548-01A1 (PI: Joe Verghese), R01AG061155-01 (PI: Sofiya Milman), and R01AG057909 (PI: Nir Barzilai). The funder had no role in the design, data collection, data analysis, and reporting of this study.

Data Availability Statement

All data generated or analyzed during this study are included in this article and its supplementary material files. Further inquiries can be directed to the corresponding author.