Age and beta amyloid deposition impact gait speed, stride length, and gait smoothness while transitioning from an even to an uneven walking surface in older adults

Abstract

Background

Capturing a measure of movement quality during a complex walking task may indicate the earliest signs of detrimental changes to the brain due to beta amyloid (Aβ) deposition and be a potential differentiator of older adults at elevated and low risk of developing Alzheimer’s disease. This study aimed to determine: 1) age-related differences in gait speed, stride length, and gait smoothness while transitioning from an even to an uneven walking surface, by comparing young adults (YA) and older adults (OA), and 2) if gait speed, stride length, and gait smoothness in OA while transitioning from an even to an uneven walking surface is influenced by the amount of Aβ deposition present in an OA’s brain.

Methods

Participants included 56 OA (>70 years of age) and 29 YA (25-35 years of age). In OA, Aβ deposition in the brain was quantified by PET imaging. All participants completed a series of cognitive assessments, a functional mobility assessment, and self-report questionnaires. Then participants performed two sets of walking trials on a custom-built walkway containing a mixture of even and uneven surface sections, including three trials with a grass uneven surface and three trials with a rocks uneven surface. Gait data were recorded using a wireless inertial measurement unit system. Stride length, gait speed, and gait smoothness (i.e., log dimensionless lumbar jerk) in the anteroposterior (AP), mediolateral (ML), and vertical (VT) directions were calculated for each stride. Outcomes were retained for five stride locations immediately surrounding the surface transition.

Results

OA exhibited slower gait (Grass: p<0.001; Rocks: p=0.006), shorter strides (Grass: p<0.001; Rocks: p=0.008), and smoother gait (Grass AP: p<0.001; Rocks AP: p=0.002; Rocks ML: p=0.02) than YA, but they also exhibited greater reductions in gait speed and stride length than YA while transitioning to the uneven grass and rocks surfaces. Within the OA group, those with greater Aβ deposition exhibited decreases in smoothness with age (Grass AP: p = 0.02; Rocks AP: p = 0.03; Grass ML: p = 0.04; Rocks ML: p = 0.03), while those with lower Aβ deposition exhibited increasing smoothness with age (Grass AP: p = 0.01; Rocks AP: p = 0.02; Grass ML: p = 0.08; Rocks ML: p = 0.07). Better functional mobility was associated with less smooth gait (Grass ML: p = 0.02; Rocks ML: p = 0.05) and with less variable gait smoothness (Grass and Rocks AP: both p = 0.04) in the OA group.

Conclusion

These results suggest that, relative to YA, OA may be adopting more cautious, compensatory gait strategies to maintain smoothness when approaching surface transitions. However, OA with greater Aβ deposition may have limited ability to adopt compensatory gait strategies to increase the smoothness of their walking as they get older because of neuropathological changes altering the sensory integration process and causing worse dynamic balance (i.e., jerkier gait). Functional mobility, in addition to age and Aβ deposition, may be an important factor of whether or not an OA chooses to employ compensatory strategies to prioritize smoothness while walking and what type of compensatory strategy an OA chooses.

1. Introduction

Alzheimer’s disease (AD) is a significant health issue because it results in a progressive loss of ability to perform activities of daily living, causing a decline in quality of life for those with AD and their caregivers (Lechowski et al., 2010; Zhu et al., 2006). Further, it results in early mortality, usually 5-12 years after the onset of symptoms (Vermunt et al., 2019). Current treatments for AD are minimally effective in slowing the progression because they can only be prescribed once symptoms of the disease are reported, after significant irreversible neuron death and brain atrophy have occurred. Significant beta amyloid (Aβ) deposition, one of the hallmarks of AD pathology, has been identified as one of the biomarkers of the preclinical phase of AD, when individuals have accrued some neurodegeneration but do not exhibit symptoms of AD (Jack et al., 2018). Thus, elevated Aβ deposition is considered to be a major risk factor for developing AD, termed by the International Working Group for New Research Criteria for the Diagnosis of AD as the “asymptomatic at risk” state (Dubois et al., 2010), and a target with potential for early prevention of AD (Dubois et al., 2016).

While Aβ deposition is most commonly associated with elevated risk of future cognitive decline and AD is most commonly associated with progressive cognitive decline (Mormino & Papp, 2018; Tarawneh & Holtzman, 2012), a number of studies have provided evidence that progression from cognitively normal to mild cognitive impairment to AD status is also associated with changes in gait and mobility. These changes in mean gait spatial and temporal metrics, such as slower gait speed and decreased stride length (Allali et al., 2016; de Melo Coelho et al., 2012; Rucco et al., 2017; Verghese et al., 2007), as well as in movement quality gait measures, such as increased gait speed variability and increased stride duration variability (Beauchet et al., 2014; Tian et al., 2018), can occur up to 12 years prior to detectable cognitive decline (Buracchio et al., 2010). However, these changes in gait are not unique to AD and are also typical of age-related declines in cognition and mobility, such as those associated with fall risk (Hausdorff et al., 1997; Imms & Edholm, 1981; Niederer et al., 2021; Verghese et al., 2002; Zukowski et al., 2020). Therefore, identifying changes in gait that can differentiate between older adults who exhibit typical age-related declines in cognition and mobility and older adults who may exhibit typical age-related declines in cognition and mobility but are also at elevated risk of developing AD based on Aβ deposition is an important aim towards developing a target with potential for early prevention of AD.

One potential for differentiating between healthy older adults at elevated and low risk of developing AD is examining gait during complex walking tasks typical of everyday life. Previous research has determined that between mean spatial and temporal gait metrics and movement quality gait measures, movement quality measures seem to be more sensitive to preclinical neurodegeneration, higher amyloid burden and cognitive decline (Tian et al., 2018), likely due to the fact that movement quality measures are associated with changes in the brain in areas that are relevant for coordination of movement and sensorimotor integration (Tian et al., 2017). Similarly, it is posited that a measure of movement quality captured during more complex gait tasks, relative to during a simple, self-selected walking speed task, may be better able to identify preclinical neuropathological changes because of the greater demand on planning and sensorimotor integration with the complex task (Tian et al. 2017). An ecologically relevant complex gait task could be transitioning from an even surface (e.g., a sidewalk) to an uneven surface (e.g., grass or rocky surface) because this is a task commonly performed in everyday life during community ambulation that requires the coordination of planning and sensorimotor integration. Thus, capturing a measure of movement quality during a complex walking task, such as transitioning from an even surface to an uneven surface, may indicate the earliest signs of detrimental changes to the brain due to Aβ deposition and be a potential differentiator of older adults at elevated and low risk of developing AD.

To date, no studies have explored how movement quality while walking on or transitioning onto uneven terrain may differ in healthy older adults versus older adults with preclinical neuropathology, cognitive decline, or mobility deficits, or on how transitioning onto uneven terrain may differ in healthy older adults versus young adults. Previous research has, however, provided evidence that young adults exhibit high levels of motor coordination when transitioning from a compliant to a firm surface during a running task by adjusting their leg stiffness to minimize vertical center of mass displacement (Ferris et al., 1999), and they increase the time spent and precision of planning upcoming foot placements as the terrain on which they are walking becomes more complex (Matthis et al., 2018). Additionally, as healthy adults walk on increasingly challenging and uneven terrain in a real-world environment, such as on gravel, grass, and woodchips, they exhibit increasing ground clearance, increasing variability of ground clearance, stride length, and gait speed, and increasing metabolic rate and cost of transport (Kowalsky et al., 2021). Further, both young and healthy older adults increase step width when walking on grass in a real-world environment, likely to increase stability, and walk more slowly on grass than asphalt (Hennah & Doumas, 2023); however, relative to young adults, healthy older adults typically exhibit less smooth (i.e., jerkier) gait while walking on an uneven brick surface (Dixon, Schütte, et al., 2018) and greater trunk center of mass acceleration variability while walking over a multi-surface terrain (Marigold & Patla, 2008). Finally, impairments in cognition, motor control, and sensory integration in people with neuropathological conditions may necessitate adopting movement strategies that differ from the movement strategies employed by healthy older adults, who must compensate for typical age-related declines in cognition, motor control, and sensory integration (Hawkins et al., 2017). Subtle impairments in cognition, motor control, and sensory integration due to elevated Aβ deposition may also elicit movement coordination strategies while transitioning onto uneven terrain that are different from those employed by healthy older adults at low risk of developing AD and thus warrants exploration.

Although prior literature has primarily focused on exploring traditional measures of movement quality, such as gait variability, in relation to preclinical neurodegeneration, other measures of movement quality, such as gait smoothness, may show neuropathological changes associated with preclinical AD earlier than traditional measures. Specifically, gait smoothness may be a more sensitive measure than gait variability because while gait variability can assess gait inconsistencies and asymmetries, gait smoothness is designed to detect subtle deviations from controlled and coordinated movements (Germanotta et al., 2022; Hogan & Sternad, 2009). There are various measures of gait smoothness that can distinguish between healthy older adults and those with dementia or other disease states (Beck et al., 2018; Ijmker & Lamoth, 2012), but log dimensionless jerk (LDLJ, i.e., an amplitude-normalized derivative of acceleration), in particular, has been shown to outperform other measures of gait smoothness in terms of ability to differentiate between young and older adults while performing a complex gait task (Dixon, Stirling, et al., 2018) and between individuals at different phases in recovery from stroke (Germanotta et al., 2022). Indeed, LDLJ has been argued to be a better measure of gait smoothness to utilize with populations who have sensorimotor deficits because the measure is not influenced by the speed of the movement, and those with sensorimotor deficits often move more slowly than unimpaired populations (Hogan & Sternad, 2009). Although to date, LDLJ has not been examined in those with preclinical neuropathology, it may be a sensitive measure of subtle deficits in sensorimotor integration and motor coordination.

Thus, gaining a better understanding of how movement quality and mean spatial and temporal gait metrics during the performance of an ecologically valid, complex everyday task are impacted by age and elevated Aβ deposition should be explored. To address this goal, this study has two aims. The first aim is to determine age-related differences in gait speed, stride length, and gait smoothness (i.e., LDLJ) while transitioning from an even to an uneven walking surface, by comparing young adults (YA) and older adults (OA), including a combined group of OA both with and without elevated Aβ deposition. Because no prior research has examined how transitioning onto uneven terrain, a very ecologically valid task, may differ in healthy older adults versus young adults, this is a very novel study aim. We hypothesize that YA will exhibit faster gait, longer strides, and smoother gait than OA. The second aim is to determine if gait speed, stride length, and gait smoothness in OA while transitioning from an even to an uneven walking surface is influenced by the amount of Aβ deposition present in an OA’s brain. This aim is significant and innovative because no studies to date have explored how gait outcomes while walking on or transitioning onto uneven terrain may differ in healthy older adults versus older adults with preclinical neuropathology. We hypothesize that higher amounts of amyloid deposition, which is related to elevated risk of developing AD, will be associated with slower gait, shorter strides and jerkier (aka, less smooth) gait.

2. Methods

2.1. Participants

Fifty-six OA and 29 YA participated in the current study. The current study involves a subset of participants from the longitudinal, observational Brain Networks and Mobility Study (B-NET, NCT03430427) that recruited individuals from Forsyth County, NC and surrounding areas. The subset of participants in the current study were recruited from those already enrolled in the parent study, with no biases in how these individuals were selected, and who were willing to participate in a supplemental visit. The inclusion criteria for the parent study were: being a community-dwelling OA over the age of 70 years or a YA between the ages of 25-35 years and able to communicate with study personnel. The exclusion criteria for the parent study were: serious or uncontrolled chronic disease, diagnosis of a neurologic disease, prior traumatic brain injury with residual effects, dependence on a walker or another person to ambulate, an unwillingness or inability to have an MRI brain scan, plans to relocate during the study period and an unwillingness to return for planned study visits, having an amputation or musculoskeletal impairments that precluded being able to perform all study procedures, participation in a structured exercise or cognitive intervention, history of brain or spinal cord tumor, significant uncorrectable hearing and/or vision impairment, and evidence of mild cognitive impairment, as determined by the study neuropsychologist based on performance on the Montreal Cognitive Assessment and detailed in our previous work (Laurienti et al., 2023). The inclusion criteria for the supplemental visit were: being able to communicate in English with study personnel, being able to walk quickly for at least two minutes at a time without assistance, and, for the OA only, a willingness to have a Aβ positron emission tomography (PET) scan as part of the parent study. Exclusion criteria for the supplemental visit were: recent surgery or hospitalization within 6 months prior to the supplemental visit and uncorrected hearing and/or vision impairments. The current study and the parent study were approved by the respective Institutional Review Boards, and all participants provided written, informed consent prior to participation in the parent and supplemental visit study procedures.

In OA, Aβ deposition in the brain was quantified by Pittsburgh Compound-B (PiB) PET imaging. More specifically, brain MRI data were acquired for all participants at Wake Forest University (WFU) on a 3T Siemens Skyra scanner using a high-resolution 32-channel head coil. MRI sequences included T1-weighted 3D volumetric MPRAGE (1x1x1mm isotropic voxels, matrix 192x240x256, TR=2300 ms, TE=2.98 ms). T1 MRI scans were processed using FreeSurfer v5.3 (https://surfer.nmr.mgh.harvard.edu) to generate target and reference regions of interest for amyloid PET scan processing. Then [11C]Pittsburgh compound-B (PIB) (Klunk et al., 2004) was used for assessing fibrillar Aβ brain deposition on PET. Participants were injected with an i.v. bolus of ~10mCi (370 MBq) (+/−10%) [11C]PiB over 5-10s, followed by 40-min uptake. A computed tomography (CT) scan was done prior to PET for attenuation correction. Emission images were acquired continuously for 40-70 min post-injection (6×5-min frames) on a 64-slice GE Discovery MI DR PET/CT scanner in the WFU PET research center. PET Images were reconstructed as multi-frame images, and motion correction was applied to PET images, which were then averaged into a 3D image. We estimated the transformation of participant CT scans to MRI space and applied the transformation to coregister PET images (in the same native space as CT) to MRI scans. A voxelwise 40-70 minute standardized uptake volume ratio (SUVR, cerebellar grey reference) image was then generated. Global brain amyloid deposition was calculated as PiB SUVR averaged from a cortical region of interest sensitive to early AD, using FreeSurfer-segmented regions (Maass et al., 2018; Mormino et al., 2012). This global PiB SUVR measure served as a biomarker of Aβ burden; a threshold of >=1.21 SUVR was used to determine Aβ+. Aβ+ was also classified using trained visual raters (Collij et al., 2021). After Aβ deposition was quantified, older adults were classified as either having elevated Aβ deposition (Aβ+, i.e., at elevated risk of developing AD) or not (Aβ−, i.e., at low risk of developing AD). OA were analyzed as a single group in both the first and second aims of this study, with Aβ deposition included in the analysis for the second aim as a continuous variable, as opposed to using a dichotomous Aβ+/Aβ− status. However, OA were divided into Aβ+/Aβ− groups for testing purposes to ensure that there were adequate numbers of OA both with and without elevated Aβ deposition and to allow for counterbalancing the testing order across the Aβ+ and Aβ− groups. PET scans and Aβ deposition quantification were completed by research personnel from the parent study. Research personnel from the parent study also assigned older adult participants to coded groups, based on their Aβ+ or Aβ− status or as an unclassified OA if they had not yet completed the PET scan, before they were scheduled to participate in the current study visit. Thus, both participants and research personnel involved in collecting and analyzing data for the current study were blinded to which participant groups were Aβ+ and Aβ−, respectively.

2.2. Procedures

As part of the parent study, demographic information, including age, gender, race/ethnicity, body mass index (BMI), and years of education completed were collected, and older adult participants completed MRI and PET scans. The timing between the PET scan and participation in the current study visit was variable due to difficulties with testing during the COVID-19 pandemic.

Participants from the OA Aβ+, OA Aβ−, and YA groups all completed the same series of procedures in the current study visit. At the start of the visit, participant height was measured. Then participants completed a series of cognitive assessments, a functional mobility assessment, and self-report questionnaires to characterize the sample. In terms of cognitive function, information processing speed was assessed using the Coding subtest from the Wechsler Adult Intelligence Scale 4^th edition (Wechsler, 2008) and executive function was assessed using the Comprehensive Trail Making Test (CTMT) (Reynolds, 2020). Functional mobility, specifically balance, gait, and repeated chair stands, was assessed using the expanded Short Physical Performance Battery (eSPPB Total) (Simonsick et al., 2001). In terms of self-report questionnaires, self-reported physical activity was assessed with the Physical Activity Scale for the Elderly (PASE) (Washburn et al., 1999), balance self-efficacy was assessed with the Activities-specific Balance Confidence Scale (ABC) (Powell & Myers, 1995), and self-perceived walking ability was assessed by the Mobility Assessment Tool—short form (MAT-sf) (Rejeski et al., 2013). Age, gender, race/ethnicity, height, BMI, years of education completed, Aβ deposition, timing between the PET scan and participation in the current study visit, cognitive function, functional mobility, and self-report measures for the YA and OA are summarized in Table 1.

Table 1. Characteristics of participants in the OA and YA groups.

Values are Mean±SD or Median(IQR).

	OA (n=55)	YA (n=29)
Demographic Characteristics
Age (years)	78.2±4.7	30.7±3.4
Gender	30 males, 25	11 males, 18
	females	females
Race/Ethnicity (counts)
Caucasian/White	49	18
African American/Black	5	4
Asian	1	3
Multiracial/Other	0	4
Height (cm)	170.6±9.6	168.1±9.6
Body mass index	28.4±3.7	25.4±5.1
Years of education completed	16 (14 – 18)	16 (16 – 18)
Aβ deposition (PiB SUVR)	1.20 (1.14 – 1.62)	n/a
Timing between PET scan and participation in current study visit (in months)	18 (7 – 19)	n/a
Cognitive Assessments
WAIS IV Coding (age-adjusted scaled score, max 19)	12 (11 – 14)	12 (11 – 14)
Comprehensive Trail Making Test (age-adjusted percentile score)	55 (34 – 77)	58 (20 – 70)
Functional Mobility Assessment
expanded Short Physical Performance Battery (Total score, max 4)	2.3±0.4	2.8±0.2
Self-reported Questionnaires
Physical Activity Scale for the Elderly	148.7±63.5	167.9±74.8
Activities-Specific Balance Confidence Scale (max. 100%)	88.8 (83.1 – 95.0)	97.5 (95.8 – 98.8)
Mobility Assessment Tool—short form	69.0 (62.0 – 70.3)	73.1 (70.5 – 73.1)

After these initial assessments, participants completed two sets of three ecologically valid walking trials on a custom-built 1-meter wide modular, wooden walkway containing a mixture of even and uneven surface sections (Figure 1). Participants completed three walking trials on the walkway containing an uneven grass surface (i.e., compliant and uneven) and three walking trials on the walkway containing an uneven rocks surface (i.e., firm and uneven) in a counterbalanced order that was the same for the OA Aβ+, OA Aβ−, and YA groups. The even surface sections were constant across all six trials. The grass surface was artificial turf layered on top of shapes of carpet foam that were then glued to the wooden surface, designed to look and feel like a lumpy grass lawn. The rocks surface was smooth river rocks that ranged in size from 1 to 5 inches in diameter and were glued down onto the wooden surface, designed to look and feel like walking across a dry riverbed. The even surface was gray grit tape applied to the wooden surface, designed to look and feel like a concrete sidewalk. For each trial, participants started 2.9 meters away from the walkway, allowing participants to take 3-5 steps to reach steady-state gait before reaching the walkway (Lindemann et al., 2008; Najafi et al., 2010), and then they walked up a short ramp with a sidewalk surface and a gradual incline, across five meters of even sidewalk surface, across five meters of an uneven surface, across two meters of even sidewalk surface, and then down another short ramp with a gradual incline and even surface. The ecological validity of this gait assessment is significant because prior research has hypothesized that more complex gait tasks, which are typical of everyday walking, may be more sensitive to subtle gait impairments due to dementia disease subtypes, relative to traditional less challenging overground walking tasks (Mc Ardle et al., 2021). Participants were instructed to “walk at their preferred, comfortable speed” at the start of each trial. To ensure participant safety, a handrail was present on the right side of the walkway, but participants were instructed to not use the handrail unless necessary to stabilize themselves. Tri-axial linear acceleration and angular velocity data were recorded using a 6-sensor wireless inertial measurement unit system (128 Hz, Opal Sensors, APDM Inc., Portland, OR). The six sensors were placed on the dorsum of each foot, on the lower back at the lumbosacral junction, on the trunk at the sternum, and on the dorsum of each wrist.

Figure 1.

Schematic of the walkway setup for the A) uneven grass and B) uneven rocks walking trials, and photos of C) the walkway with the grass panels installed and the rock panels in the background, D) a close-up of the rocks surface, and E) a close-up of the grass surface.

Stride lengths were determined by using double integration of the foot acceleration signals, following a zero-velocity update and orientation correction as described by Rebula et al. (2013). Gait speed was calculated on a per-stride basis using the stride length divided by the stride time. Gait smoothness was calculated using the LDLJ of the trunk acceleration in each direction (anteroposterior, AP; mediolateral, ML; vertical, VT) for each stride (Balasubramanian et al., 2015; Melendez-Calderon et al., 2020), with more positive (or less negative) LDLJ values indicating smoother gait. Outcomes were retained for five stride locations immediately surrounding the surface transition: two strides before the transition (Approach stride); one stride before the transition (Pre-Transition stride); the Transition stride; one stride after the transition (Post-Transition stride); and two strides after the transition (Uneven stride). For all trials, the Transition stride was defined based on the first step on the grass or rock surface (step n), where the Transition stride covered heel contact of step n−1 to step n+1. (Figure 2). Average values for the five stride locations were calculated across all three rocks trials and across all three grass trials. All analysis was completed using custom scripts in MATLAB (MathWorks, Natick, MA).

Figure 2.

Depiction of the Approach, Pre-Transition, Transition, Post-Transition, and Uneven Strides on the walkway, in the case that the first step onto the uneven section was a Right foot placement.

2.3. Statistical Analysis

There were a few instances of missing data. One OA was excluded from all analyses due to technical difficulties, resulting in 55 OA in the analyzed sample. In the OA group, one participant opted against undergoing a PET scan after completing the current study visit and was thus not classified as Aβ+ or Aβ−; this participant was excluded from the analysis in the second aim. Of the remaining OA, 22 were classified as Aβ+ and 32 were classified as Aβ−. Only two of the three grass trials were analyzed for one OA (Aβ−) because the participant misinterpreted the task instructions during the first trial. Additionally, two OA did not complete the Coding subtest due to hand tremors, not due to a neurological disease, that would have invalidated the completion time used to score the test. In the YA group, one rocks trial was excluded for one participant and all three rocks trials were excluded for another participant, across the analyses for both study aims, because the individuals took too long of strides to capture an Approach and a Pre-Transition stride before the Transition stride on the even sidewalk surface. Additionally, due to technical difficulties, one YA did not complete two of the three grass trials or the MAT-sf.

For the first aim, to test the effects of age group on stride length, gait speed, LDLJ-AP, LDLJ-ML, and LDLJ-VT as participants crossed from an even to an uneven surface, linear mixed-effects models were fit for each outcome. Models were stratified by condition, with separate models for the rocks and grass conditions. Each model contained fixed effects of stride location (Approach, Pre-Transition, Transition, Post-Transition, and Uneven), age group (YA and OA), and their interaction. Random intercepts by participant and random slopes by participants across the three trials for each condition were included in the models to account for within-subject correlations of the fifteen repeated measures (5 stride locations x 3 trials) in each condition. Final model selection using random slopes versus just random intercepts was confirmed by comparing AIC; random slopes and random intercepts had lower AIC in all cases. Additionally, gait speed was included as a covariate in all LDLJ models. Assumptions of normality were confirmed by examining residual plots of each model. Type 3 tests for fixed effects (i.e., F-Test) were first interpreted for each model to assess overall differences between stride locations, groups, and their interaction. Significant stride location*group interaction effects (α=0.05) were further examined using pairwise contrasts. A secondary, post-hoc analysis compared the variability of LDLJ values, within each direction, between age groups using independent sample t-tests. The variability was calculated using the standard deviation of LDLJ values across all 15 stride locations (5 stride locations x 3 trials) in each condition.

For the second aim, to test the effects of age in OA, Aβ deposition and uneven condition on stride length, gait speed, LDLJ-AP, LDLJ-ML, and LDLJ-VT, similar linear mixed-effects models were fit for each outcome, stratified by condition (e.g., rocks vs. grass). Each model included fixed effects of stride location, age (continuous), Aβ deposition, and the interaction of age and Aβ deposition. Random intercepts by participant and random slopes by participants across trials were included in the models to account for within-subject correlations. Gait speed was included as a covariate in all LDLJ models, and time between PET scan and the current study visit was included as a covariate in all models. Type 3 tests for fixed effects assessed the overall effect of stride location, age, Aβ deposition, and the interaction of age and Aβ deposition for the rocks and grass conditions. Significant age*Aβ deposition interactions were further examined using pair-wise contrasts.

Finally, to determine how gait smoothness during a surface transition task in OA is associated with standardized clinical assessments, we conducted an exploratory, post-hoc correlational analysis. Pearson correlation coefficients explored the univariate relationship between the average LDLJ-AP and LDLJ-ML values at the Transition stride for each condition (grass, rocks) and the average and variability of LDLJ-AP and LDLJ-ML across all strides within a condition with Aβ deposition, functional mobility (eSPPB Total), cognitive function (Coding and CTMT), and self-report measures (PASE, ABC, and MAT-SF) in OA. Only gait smoothness in the AP and ML directions were included in this analysis to minimize the number of additional exploratory analyses and to prioritize the analysis of these novel gait outcomes.

A 0.05 significance level was used throughout, and all statistical analyses were conducted in MATLAB R2020a using the Statistics and Machine Learning Toolbox.

3. Results

3.1. Effects of age group and uneven walkway condition on stride length and gait speed

For stride length, there was a main effect of age group during the grass (p=0.028) and rocks (p=0.013) conditions, with OA taking 0.08 m and 0.09 m shorter strides than YA during the Approach stride, respectively (Table 2). There was also a main effect of stride location (p=0.019) and a group by stride location interaction (p<0.001) in the grass condition (Figure 3, Table 2), where pair-wise contrasts indicated that YA took a 0.02 m longer Post-Transition stride (p=0.035), but OA took 0.05 m shorter Post-Transition (p<0.001) and Uneven (p<0.001) strides relative to the Approach stride. In the rocks condition, pair-wise contrasts indicated that YA did not change stride length across the five stride locations (all p>0.05, Figure 4, Table 2), but OA reduced stride length (group by stride location interaction p=0.008, Table 2) during the Transition (−0.03 m, p=0.03), Post-Transition (−0.04 m, p=0.001), and Uneven (−0.03 m, p=0.03) strides, relative to the lack of change in stride lengths exhibited by YA.

Table 2. Average stride length, gait speed, and LDLJ during each of the five stride locations and averaged across the three trials during the grass and rocks conditions in YA and OA.

Values are Mean(SD).

	Approach	Pre- Transition	Transition	Post- Transition	Uneven	Stride Location p value (F-test)	Group p value (F test)	Group*Stride Location p value (F Test)
Grass
Stride Length (m)
YA	1.39(0.12)	1.38(0.13)	1.38(0.14)	1.41(0.13)	1.40(0.14)	0.02 *	0.03 *	< 0.001 *
OA	1.30(0.15)	1.29(0.16)	1.28(0.18)	1.25(0.18)	1.25(0.19)
Gait Speed (m/s)
YA	1.28(0.12)	1.28(0.14)	1.25(0.14)	1.25(0.14)	1.24(0.13)	< 0.001 *	0.10	< 0.001 *
OA	1.21(0.20)	1.17(0.20)	1.16(0.23)	1.11(0.22)	1.11(0.22)
LDLJ – AP (no units)
YA	−4.98(0.57)	−5.08(0.67)	−4.83(0.55)	−4.79(0.55)	−4.72(0.49)	< 0.001 *	< 0.001 *	< 0.001 *
OA	−4.39(0.46)	−4.42(0.46)	−4.37(0.39)	−4.37(0.32)	−4.38(0.38)
LDLJ – ML (no units)
YA	−4.57(0.26)	−4.62(0.40)	−4.60(0.30)	−4.59(0.27)	−4.59(0.31)	0.80	0.35	0.58
OA	−4.49(0.32)	−4.46(0.27)	−4.46(0.39)	−4.47(0.33)	−4.48(0.34)
LDLJ – VT (no units)
YA	−2.48(0.27)	−2.53(0.30)	−2.54(0.26)	−2.55(0.22)	−2.51(0.24)	0.02 *	0.61	0.09
OA	−2.47(0.29)	−2.48(0.29)	−2.44(0.31)	−2.47(0.33)	−2.46(0.29)
Rocks
Stride Length (m)
YA	1.38(0.13)	1.38(0.14)	1.36(0.13)	1.36(0.15)	1.37(0.14)	0.35	0.01 *	0.008 *
OA	1.29(0.15)	1.28(0.16)	1.24(0.18)	1.23(0.19)	1.24(0.19)
Gait Speed (m/s)
YA	1.27(0.14)	1.25(0.16)	1.21(0.16)	1.24(0.16)	1.24(0.15)	< 0.001 *	0.13	0.006 *
OA	1.19(0.20)	1.15(0.21)	1.09(0.26)	1.12(0.26)	1.11(0.23)
LDLJ – AP (no units)
YA	−4.96(0.61)	−5.08(0.60)	−5.12(0.63)	−5.17(0.69)	−5.13(0.61)	< 0.001 *	< 0.001 *	0.002 *
OA	−4.43(0.43)	−4.40(0.42)	−4.64(0.40)	−4.67(0.38)	−4.67(0.43)
LDLJ – ML (no units)
YA	−4.60(0.35)	−4.68(0.40)	−4.68(0.35)	−4.77(0.32)	−4.76(0.39)	< 0.001 *	0.04 *	0.02 *
OA	−4.44(0.32)	−4.45(0.34)	−4.60(0.36)	−4.68(0.35)	−4.61(0.34)
LDLJ – VT (no units)
YA	−2.52(0.30)	−2.55(0.31)	−2.77(0.29)	−2.82(0.26)	−2.80(0.23)	< 0.001 *	0.79	0.80
OA	−2.45(0.28)	−2.46(0.30)	−2.67(0.34)	−2.73(0.31)	−2.74(0.28)

Note. Significant p-values are bolded and denoted with an *.

Figure 3.

Violin plots for each gait outcome in the grass condition across the five stride locations (Approach, Pre-Transition, Transition, Post-Transition, and Uneven) for the young adult (YA; gray) and older adult (OA; blue) groups. Violin plots illustrate the distribution of data along with the median (circle), mean (horizontal line) and interquartile range (vertical bar). Overall age effects are indicated with a * symbol for each outcome. Symbols over specific strides indicate a significant effect of stride (i.e., a significant difference relative to the Approach stride in young adults; +) or a significant group-by-stride interaction (i.e., a significant difference between the YA and OA change relative to the Approach stride; ^).

Figure 4.

Violin plots for each gait outcome in the rocks condition across the five stride locations (Approach, Pre-Transition, Transition, Post-Transition, and Uneven) for the young adult (YA; gray) and older adult (OA; blue) groups. Violin plots illustrate the distribution of data along with the median (circle), mean (horizontal line) and interquartile range (vertical bar). Overall age effects are indicated with a * symbol for each outcome. Symbols over specific strides indicate a significant effect of stride (i.e., a significant difference relative to the Approach stride in young adults; +) or a significant group-by-stride interaction (i.e., a significant difference between the YA and OA change relative to the Approach stride; ^).

Gait speed did not differ across groups during the Approach stride during either condition (both p>0.05), but there was a main effect of stride location in the grass and rocks conditions (both p<0.001) and group by stride location interactions (Grass: p<0.001; Rocks: p=0.006), indicating speed slowed during and after the transition, with OA slowing down more than YA (Table 2). Relative to the Approach stride, pair-wise contrasts indicated that YA slowed during both the grass and rocks conditions on the Transition (Grass: −0.02 m/s, p=0.03; Rocks: −0.06 m/s, p<0.001), Post-Transition (Grass: −0.03 m/s, p=0.002; Rocks: −0.03 m/s, p=0.01), and Uneven (Grass: −0.03 m/s, p=0.001; Rocks: −0.03 m/s, p=0.01) strides. Compared to changes in gait speed of YA, pairwise contrasts indicated that OA slowed down more on grass during the Pre-Transition (−0.4 m/s, p=0.001), Transition (−0.03 m/s, p=0.04), Post-Transition (−0.07 m/s, p<0.001), and Uneven (−0.07 m/s, p<0.001) strides. OA also slowed down more than YA on rocks during the Transition (−0.04 m/s, p=0.01), Post-Transition (−0.04 m/s, p=0.01), and Uneven (−0.05 m/s, p<0.001) strides.

3.2. Effects of age group and uneven walkway condition on LDLJ-AP, LDLJ-ML, and LDLJ-VT

For smoothness in the AP direction, there was a main effect of age group during the grass and rocks conditions (both p < 0.001), with OA exhibiting smoother gait (both p < 0.001) than YA in both conditions during the Approach stride (Table 2). There was also a main effect of stride location during the grass and rocks conditions (both p < 0.001) and a group by stride location interaction (Grass: p < 0.001; Rocks: p = 0.002) that was driven by changes in the YA group across stride locations (Table 2). In the grass condition, pair-wise contrasts indicated that YA walked with jerkier (aka, less smooth) gait during the Pre-Transition stride (p = 0.02), and smoother gait during the Transition (p = 0.001), Post-Transition (p < 0.001), and Uneven (p < 0.001) strides, relative to the Approach stride. In contrast, OAs did not vary smoothness across any stride location relative to the Approach stride (all post-hoc contrast p > 0.05). In the rocks condition, YA walked with jerkier gait during the Pre-Transition (p = 0.03), Transition (p = 0.002), Post-Transition (p < 0.001), and Uneven (p = 0.001) strides, relative to the Approach stride. In contrast, OAs did not change smoothness during the Pre-Transition stride relative to the Approach stride (p > 0.05) but exhibited similar decreases in smoothness as YA across all other stride locations (all other contrasts p < 0.001). Gait speed was not a significant covariate during either condition (both p > 0.05). Overall, relative to YA, OA exhibited less variability in LDLJ-AP in grass conditions (95% CI of difference = [0.04, 0.13]; p < 0.001), but not in rocks conditions (95% CI of difference = [−0.08, 0.02]; p > 0.05)

In the ML direction, smoothness did not vary by age group or stride location during the grass condition (both p > 0.05). Gait speed was a significant covariate of smoothness though, with participants exhibiting jerkier gait as gait speed increased (p = 0.02). During the rocks condition, there was a main effect of age group (p = 0.04), with OAs exhibiting smoother Approach strides compared to YA. There was also a main effect of stride location (p < 0.001) and a group by stride location interaction (p = 0.02), such that both groups decreased smoothness during the Post-Transition (p < 0.001) and Uneven (p < 0.001) strides. The group by stride location interaction was driven by consistent, but not statistically significant, patterns where pair-wise contrasts indicated that OA exhibited further decreases in smoothness, relative to YA, during the Transition (p = 0.07) and Post-Transition (p = 0.10) strides. Gait speed was not a significant covariate (p > 0.05). Overall, relative to YA, OA exhibited greater variability in LDLJ-ML in rocks conditions (95% CI of difference = [0.01, 0.06]; p = 0.03), but not in grass conditions (95% CI of difference = [−0.02, 0.04]; p > 0.05).

In the VT direction, there was a main effect of stride location during the grass (p = 0.02) and rocks (p < 0.001) conditions, with smoothness decreasing during the Transition (Grass: p = 0.002; Rocks: p < 0.001), Post-Transition (Grass: p = 0.003; Rocks: p < 0.001), and Uneven (Grass: p = 0.03; Rocks: p < 0.001) strides in YA after adjusting for gait speed (significant covariate: both p < 0.001). Pair-wise contrasts indicated that OA exhibited similar smoothness values as YA during the Approach stride and similar decreases in smoothness to YA across the other stride locations, as evidenced by no main effect of age group and no age group by stride location interaction during either condition (all p > 0.05). Overall, YA and OA did not exhibit different LDLJ-VT variability in grass (95% CI of difference = [−0.01, 0.05]; p > 0.05) or rocks conditions (95% CI of difference = [−0.02, 0.06]; p > 0.05).

3.3. Effects of age, Aβ deposition, and uneven walkway condition on stride length and gait speed

Within the OA group, there was no effect of age (as a continuous variable), Aβ deposition, or their interaction on stride length or gait speed (Figure 5) across any condition when adjusting for stride location (all p > 0.05).

Figure 5.

Scatter plots of gait speed over each stride location for the grass (top) and rocks (bottom) conditions. Each data point represents a single trial from one participant (e.g., each participant is represented 3 times per plot). The color of each data point represents that participant’s Aβ deposition as indicated by the PiB SUVR, with young adult (YA) participants represented in black. The mean of the YA group is depicted in the solid black line. Trendlines in the older adult (OA) group are depicted in cyan for more-certain Aβ− visual ratings (i.e., Aβ deposition < 1.21 global PiB SUVR) and in magenta for more-certain Aβ+ visual ratings (i.e., Aβ deposition > 1.4 PiB SUVR). Individuals initially classified as exhibiting intermediate Aβ deposition (i.e., 1.21-1.4 PiB SUVR), where trained visual rating results were mixed, were included as scatter plot data points but were not included in the calculation of the OA trendlines.

3.4. Effects of age, Aβ deposition, and uneven walkway condition on LDLJ-AP, LDLJ-ML, and LDLJ-VT

Within the OA group during both the grass and rocks conditions, there was a main effect of age in the AP direction (Grass: Beta = 0.09, SE = 0.04, p = 0.01; Rocks: Beta = 0.09, SE = 0.04, p = 0.02) and a similar, but non-significant, effect in the ML direction (Grass: Beta = 0.05, SE = 0.03, p = 0.08; Rocks: Beta = 0.05, SE = 0.03, p = 0.07, Figure 6). Specifically, consistent with the YA vs. OA comparison, increased age in those with relatively lower Aβ deposition was associated with smoother AP trunk motion compared to younger OA with similar, but statistically non-significant increases in ML smoothness with increasing age. During both grass and rocks conditions, there was also a main effect of Aβ deposition in the AP (Grass: Beta = 4.59, SE = 1.94, p = 0.02; Rocks: Beta = 4.50, SE = 1.98, p = 0.02) and ML directions (Grass: Beta = 3.22, SE = 1.52, p = 0.03; Rocks: Beta = 3.38, SE = 1.58, p = 0.03), with Aβ deposition positively associated with smoothness in younger OA in AP and ML directions. The interaction between age and Aβ deposition was significant during the grass and rocks conditions in the AP (Grass: Beta = −0.06, SE = 0.02, p = 0.02; Rocks: Beta = −0.05, SE = 0.02, p = 0.03) and ML directions (Grass: Beta = −0.04, SE = 0.02, p = 0.04; Rocks: Beta = −0.04, SE = 0.02, p = 0.03). These interactions indicate that smoothness decreased with age in those with elevated Aβ deposition across both conditions and in both AP and ML directions (Figure 6). Taken together, these results provide evidence that older adults with relatively low Aβ deposition exhibited increasing smoothness with increasing age, but older adults with relatively high Aβ deposition exhibited decreasing smoothness with increasing age. Gait speed was a significant covariate in the ML direction for the grass condition (Beta = −0.30, SE = 0.10, p = 0.004) but not the rocks condition or in either condition in the AP direction (all p > 0.05). There were no main effects of age or Aβ deposition, and there was no age by Aβ deposition interaction effect in the VT direction during the grass and rocks conditions (all p > 0.05). Gait speed, however, was a significant covariate during the grass and rocks conditions (Grass: Beta = −0.52, SE = 0.08, p < 0.001; Rocks: Beta = −0.46, SE = 0.08, p < 0.001)).

Figure 6.

Scatter plots of gait smoothness (i.e., LDLJ) in the AP direction, by age for each stride location for the grass (top) and rocks (bottom) conditions. Each data point represents a single trial from one participant (e.g., each participant is represented 3 times per plot). The color of each data point represents that participant’s Aβ deposition as indicated by the PiB SUVR, with young adult (YA) participants represented in black. The mean of the YA group is depicted in the solid black line. Trendlines in the older adult (OA) group are depicted in cyan for more-certain Aβ− visual ratings (i.e., Aβ deposition < 1.21 global PiB SUVR) and in magenta for more-certain Aβ+ visual ratings (i.e., Aβ deposition > 1.4 PiB SUVR). Individuals initially classified as exhibiting intermediate Aβ deposition (i.e., 1.21-1.4 PiB SUVR), where trained visual rating results were mixed, were included as scatter plot data points but were not included in the calculation of the OA trendlines.

3.5. Relationship between LDLJ-AP and LDLJ-ML and functional mobility, cognitive function, and self-reported measures

Within the OA group, greater within-subject inter-stride variability in LDLJ-AP on grass was associated with worse information processing speed, as indicated by lower Coding scores (r = −0.31, p = 0.02), and worse functional mobility, as indicated by lower eSPPB Total scores (r = −0.28, p = 0.04). On rocks, greater variability in LDLJ-AP was associated with worse functional mobility, as indicated by lower eSPPB Total scores (r = −0.28, p = 0.04). Additionally, smoother LDLJ-AP during the Transition stride on the rocks was associated with worse balance confidence, as indicated by lower ABC scores (r = −0.28, p = 0.04). There were no other associations between LDLJ-AP (mean, variability of, or during the Transition stride) with any of the other measures (all other p > 0.05). In the ML direction, smoother average LDLJ-ML and LDLJ-ML during the Transition stride on grass were associated with worse functional mobility, as indicated by lower eSPPB Total scores (Average: r = −0.32, p = 0.02; Transition: r = −0.33, p = 0.01). On the rocks, smoother average LDLJ-ML was associated with worse functional mobility, as indicated by lower eSPPB Total scores (r = −0.28, p = 0.05). There were no other associations between LDLJ-ML (mean, variability of, or during the Transition stride) with any of the other measures (p > 0.05).

4. Discussion

The purpose of this study was to determine age-related differences in gait speed, stride length, and LDLJ in three directions while transitioning from an even to an uneven walking surface, by comparing young adults (YA) and older adults (OA), and to determine if these same gait metrics were influenced by the amount of Aβ deposition present in an older adult’s brain. This study has three main findings. The first main finding is that OA exhibited greater changes in gait speed and stride length when walking over transitions compared to YA, but YA exhibited greater changes in smoothness. The second main finding is that, within the OA group, those with greater Aβ deposition exhibited decreases in smoothness with age, while those with lower Aβ deposition exhibited increasing smoothness with age, revealing a potential breakdown in age-related compensation for those with greater Aβ deposition. The third main finding is that better functional mobility was associated with less smooth gait and with less variable gait smoothness in the OA group.

In partial support of our hypothesis, YA exhibited faster gait and longer strides than OA, but YA did not exhibit smoother gait than OA. Indeed, interestingly, OA exhibited smoother gait in the AP direction than YA during the majority of the stride locations in the rocks and grass conditions and largely did not differ in smoothness, relative to YA, in the ML and VT directions during the rocks and grass conditions. The smoother gait of OA is especially notable because OA not only exhibited slower gait and shorter strides than YA, but they also exhibited greater reductions in gait speed and stride length than YA while transitioning onto and then walking on the uneven grass and rocks surfaces. The slower gait and shorter stride lengths of OA are in agreement with previous literature that determined OA, relative to YA, take shorter steps, resulting in slower gait, as a result of lower ankle plantarflexor strength and power (JudgeRoy et al., 1996). These authors further determined that OA generate near maximal ankle plantarflexor power while walking at their self-selected speed (JudgeRoy et al., 1996), which would limit their ability to easily adapt to external perturbations and has been associated with increased likelihood of falling (Sieri & Beretta, 2004). Although previous literature provides evidence that OA walk with less smoothness on an uneven brick surface, relative to YA (Dixon, Schütte, et al., 2018), this previous analysis focused on walking on an uneven surface, as opposed to transitioning from an even to an uneven walking surface, which was the novel focus of the current study. Importantly, prior work examining gait fluidity during stair approach and descent has reported similar findings to the current study. Specifically, OA exhibit more fluid (i.e., smooth) AP motion and slower gait speeds compared to YA when approaching stairs (Telonio et al., 2014). Combined, these results suggest that OA may be adopting more cautious, compensatory gait strategies (e.g., shorter, slower strides) to maintain smoothness when approaching surface transitions (Telonio et al., 2014), potentially to minimize likelihood of falling.

The association between age, Aβ deposition, and smoothness suggests these compensatory gait strategies, such as prioritizing gait smoothness, may be dependent on the availability of cognitive resources in OAs, and the effectiveness of these strategies over time may depend on an individual’s Aβ accumulation. While we did not observe a relationship between Aβ deposition and gait speed or stride length, OA with lower Aβ deposition exhibited greater smoothness with age, consistent with an increasing use of more cautious, compensatory gait strategies with increasing age (Herssens et al., 2018; Telonio et al., 2014). However, OA with greater Aβ deposition exhibited decreases in smoothness with increasing age. Thus, the combination of older age and greater Aβ deposition may disrupt the ability to adopt compensatory gait strategies. Prior work examining stair descent similarly reported worse gait fluidity (i.e., smoothness) during a stair descent task in OAs with mild cognitive impairment (MCI) compared to OAs without MCI (Charette et al., 2020). Further, these differences in gait fluidity existed despite no differences in gait speed, aligning with the results of the current study. Compensatory gait strategies may require additional cognitive resources to plan movements and integrate sensorimotor information that may be limited in people experiencing, or at risk for, cognitive decline due to preclinical neuropathological changes associated with elevated Aβ deposition (Charette et al., 2020; Telonio et al., 2014; Tian et al., 2017). Indeed, Yogev-Seligmann et al. (2022) determined that when individuals with a neuropathological condition utilized a cognitive movement strategy, specifically focusing attention on increasing step length, they walked faster and took longer steps, but the authors also observed a reduction in brain engagement index, as assessed by frontal lobe EEG, after two minutes of walking, with the reduction maintained for the subsequent eight minutes of walking. In Parkinson’s disease, these results suggest that the cognitive movement strategy resulted in higher attentional demands, as compared to typical walking, and a limited ability to sustain increased attention on this type of movement strategy (Agosta et al., 2020; Yogev-Seligmann et al., 2022; Zgaljardic et al., 2003). It is possible that OA with greater Aβ deposition may similarly have limited ability to adopt or maintain compensatory gait strategies to increase the smoothness of their walking as they get older. Further, prior analyses of the relationship between postural control and cognitive decline in OA determined that while OA without cognitive impairment exhibited active, velocity-based control of their center of mass to maintain balance, those with mild cognitive impairment exhibited worse, lower-order (i.e., position-based) control (Deschamps et al., 2014). Thus, the jerkier gait exhibited by OA with greater Aβ deposition, as age increases, may be the result of these neuropathological changes altering the sensory integration process and disallowing these individuals to utilize velocity-based balance control, causing worse dynamic balance (i.e., jerkier gait) (Deschamps et al., 2014). While the consequences of decreased gait smoothness remain unclear, the consistent results across different transition tasks suggest OA at risk for cognitive decline may have impaired gait adaptability and subsequently be at increased risk of falling (Caetano et al., 2016; Pottorf et al., 2022).

The association between eSPPB performance, gait smoothness, and variability of gait smoothness suggests that functional mobility, in addition to age and Aβ deposition, is an important factor of whether or not an OA chooses to employ compensatory strategies to prioritize smoothness while walking and what type of compensatory strategy an OA chooses. In the OA group, better performance on the eSPPB was associated with less smooth gait in the ML direction but less variability of gait smoothness in the AP direction. In other words, OA with worse functional mobility may prioritize walking with smoother gait, while those with better functional mobility may prioritize walking with less variability of gait smoothness (i.e., a more consistent level of gait smoothness), instead of walking with smoother gait. These results are in agreement with previous work that observed OA to exhibit smoother gait but a slower maximal walking speed and worse balance than YA (Telonio et al., 2014). Interestingly, worse information processing speed, as assessed by the Coding subtest, was associated with greater variability of smoothness in the AP direction on the grass. These results are in agreement with previous literature that observed an association between worse information processing speed and increasing gait variability (Martin et al., 2013). OA with faster information processing speed and better functional mobility may be less concerned about walking with smooth gait because they are able to adapt quickly to the environment, as evidenced by their lower gait smoothness variability. In contrast, OA with slower information processing speed and worse functional mobility may prioritize walking with smooth gait because they are less able to adapt to the environment as needed.

5. Limitations

This study has two main limitations. First, the timing between the PET scan and the current study visit was variable due to challenges with testing during the COVID-19 pandemic. We believe that the impact of the variable timing difference was negligible because Aβ deposition occurs slowly, approximately at the rate of −0.02 to +0.07 PiB SUVR per year in OA who are Aβ− or Aβ+ but without cognitive impairment (Hanseeuw et al., 2019). Additionally, all OA in the current study were determined to be without cognitive impairment at the start of the parent study and completed the current study visit within 24 months of their PET scan. However, we adjusted for the timing differences in the analysis of the OA group for the second study aim to ensure there was no impact of the variable timing difference. None of the timing difference covariates were significant in any of the models for the second aim analysis, confirming that the timing differences did not impact the results. The timing differences could not be added as a covariate in the analyses for the first study aim, as there were no values that could be entered for the YA group because they did not complete a PET scan. Second, this study is limited by a relatively small sample size in relation to the number of analyses. The original analysis plan was to compare stride length, gait speed, and LDLJ between the OA Aβ+, OA Aβ−, and YA groups in a single analysis, but this analysis had to be adjusted in order to account for the difficulties with testing due to COVID-19 that resulted in variable time differences across the OA group. Instead of one single analysis, we compared the gait metrics between OA and YA in one analysis and then determined the impact of Aβ deposition on those same gait metrics in OA in a second analysis, adjusting for the time differences between the PET scan and the current study visit. However, the novel and ecologically valid experimental design of capturing transitions from even to uneven surfaces and unique study sample, which included those at elevated and low risk of developing AD based on Aβ deposition, makes this study an important first step to understand how age and Aβ deposition impact performance of complex everyday tasks. Some of the surprising and novel results observed in this study underscore the need for more research that involves the performance of complex and ecologically valid walking tasks to better identify subtle gait impairments due to age-related declines and dementia disease subtypes. Subsequent studies should include a larger sample size.

6. Conclusions

The purpose of this study was to determine if YA and OA differ in gait speed, stride length, and gait smoothness while transitioning from an even to an uneven surface, and if these same gait metrics are influenced by Aβ deposition present in the brain in OA. OA exhibited smoother gait, slower gait, shorter strides, and greater reductions in gait speed and stride length than YA while transitioning onto and then walking on the uneven surfaces. These results suggest that OA are utilizing a more cautious gait strategy to maintain smoothness when transitioning from one surface type to another. Additionally, OA with greater Aβ deposition exhibited less gait smoothness with age, while those with lower Aβ deposition exhibited greater gait smoothness with age. Thus, OA with greater Aβ deposition may have fewer cognitive resources available to adopt compensatory gait strategies while transitioning onto uneven terrain, potentially putting them at increased risk of falling. Gait smoothness while transitioning from an even to an uneven surface should be explored further as a potential clinical tool for early identification of elevated risk of developing AD.

Highlights.

• Age and Aβ deposition impact performance of transitioning to an uneven surface

• Older adults exhibited more cautious gait strategies than young adults

• Older adults exhibited smoother gait when approaching transitions than young adults

• Greater Aβ deposition may result in a breakdown in age-related gait compensation

Data availability

Data will be made available on request.