Sensorimotor function as an early marker of cognitive decline and alzheimer’s biomarker burden

Abstract

Early identification of markers of cognitive decline in cognitively normal older adults is essential for dementia prevention strategies. Sensorimotor measures, such as gait speed, grip strength, and reaction time, may provide sensitive indicators of current and future impairment. This study examined associations between baseline sensorimotor function and cognitive decline in cognitively normal older adults. In this prospective cohort, 246 cognitively normal older adults from the DRIVES Project completed baseline assessments of grip strength, gait speed, simple reaction time, and the Preclinical Alzheimer Cognitive Composite (PACC). Participants were followed for a mean of 4 years. Linear mixed-effects models adjusted for demographics, APOE ε4 status, and neighborhood deprivation. Cross-sectional analyses evaluated the associations between sensorimotor measures and cerebrospinal fluid (CSF) biomarkers of amyloid, tau, and plasma neurofilament light chain (NfL). Participants (mean age: 74.9 ± 5.17 years; 48.8% female) had a mean baseline PACC score of 1.06 ± 0.50. Cross-sectionally, slower gait speed was associated with higher CSF tau/Aβ₄₂ ratio (p = 0.027), CSF tTau/Aβ₄₂ ratio (p = 0.009), and plasma NfL (p = 0.024). A slower reaction time was associated with lower baseline PACC scores (p = 0.028). Longitudinally, low grip strength (p = 0.002) and slow gait speed (p < 0.001) were predictive of lower cognitive performance, with slow gait speed also predicting a faster decline (p = 0.015). Sensorimotor function measures are associated with current and future cognitive performance, supporting their role in early identification of older adults at risk for cognitive decline.

Graphical Abstract

Introduction

Dementia has emerged as a critical public health challenge in the global aging population. Worldwide, over 57 million people were estimated to be living with dementia in 2021, a figure projected to rise to nearly 150 million by 2050 [1]. Alzheimer’s disease (AD) accounts for 60–70% of dementia cases and affects an estimated 7.2 million people in the U.S. aged 65 and older in 2025 [2]. With no cure, the emphasis has shifted to early detection during the preclinical stage, when pathology accumulates before symptoms appear. Research criteria now recognize a prolonged preclinical phase of AD, in which pathological brain changes accumulate for years before overt cognitive symptoms [3, 4]. Identifying sensitive markers of early cognitive decline during this asymptomatic stage is therefore a priority [4]. Composite neuropsychological measures have been developed for this purpose, where the Preclinical Alzheimer Cognitive Composite (PACC) has been shown to detect subtle amyloid-related cognitive decline in clinically normal older adults [3, 5, 6]. These cognitive composites, which integrate memory and executive function tests, can reveal very mild cognitive deficits that may herald future impairment. There is also growing interest in brief functional markers of brain health in older adults, particularly measures of physical performance, which might reflect early neurodegenerative changes [7–10].

Motor function, especially gait, has emerged as a promising indicator of brain aging and neurodegeneration. One study found that cognitively normal older adults at baseline who developed impairment or dementia later were walking significantly more slowly 3 to 5 years before diagnosis [8]. These findings support the concept of the “motoric cognitive risk” syndrome [11], a pre-dementia condition defined by slow gait and subjective cognitive complaint, which is prevalent in ~ 10% of older adults and predicts progression to significant cognitive decline [7]. The syndrome is operationalized to exclude individuals with Parkinson’s disease, stroke, or other motor disorders affecting the cerebellum or basal ganglia, ensuring its specificity as a potential prodromal stage of cognitive decline rather than a marker of broader neurological impairment [7, 11]. Gait speed is sometimes referred to as the “sixth vital sign” of geriatric health, as it integrates multi-system functioning, including brain circuitry involved in mobility [7]. As such, detecting subtle gait impairment in an otherwise high-functioning older person could raise concern for incipient cognitive decline, even if standard cognitive tests are still normal.

Grip strength, a simple indicator of overall muscle strength, has garnered attention as a potential marker of brain health. Loss of grip strength is a recognized manifestation of age-related sarcopenia and frailty [9]. Epidemiological studies consistently find that older adults with weaker handgrip strength tend to have worse cognitive function and a higher likelihood of future cognitive impairment [12, 13]. Mechanistically, several pathways may link reduced peripheral muscle strength to declining cognition, including chronic inflammation, hormonal changes, and shared degenerative processes that affect both the muscle and nervous systems [14].

Another functional domain that may signal emerging cognitive problems is psychomotor speed, often measured by simple reaction time tasks. Processing speed typically declines with normal aging and even more so in neurodegenerative conditions [10, 15]. Slower and more variable reaction times have been observed in older adults with mild cognitive impairment and early dementia compared to cognitively healthy peers [16]. Such reaction time measures have predictive value in a longitudinal study; greater reaction time variability at baseline nearly doubled the risk of transitioning to mild cognitive impairment within 5 years [17]. Impairments in physical capability and cognition often co-occur in older individuals, and converging evidence points to a common biological substrate underlying this dual decline. The concept of “cognitive frailty” has been introduced to describe older adults with simultaneous physical frailty and mild cognitive impairment in the absence of dementia. This condition carries a high risk for progression to AD [7, 18, 19]. Consistent with this, one community study found that participants with both significant gait slowing and memory decline had over six times greater risk of developing dementia than those without either decline [20]. Such “dual decliners” likely reflect advanced brain pathological changes affecting widespread networks involved in both mobility and cognition [21]. Altogether, these observations underscore that late-life motor slowing, physical frailty, and cognitive impairment are deeply interrelated phenomena, often mediated by the same neurological insults.

These sensorimotor functions not only predict future cognitive decline but also mirror underlying Alzheimer’s disease (AD) pathology. Individuals with stronger grip strength and faster walking speeds have significantly lower plasma concentrations of neurodegeneration markers [9]. Likewise, physically frail older adults have elevated circulating levels of phosphorylated tau compared to their non-frail peers, suggesting that tau pathology may underlie the association between frailty and cognition [19]. Gait speed impairments have also been tied to central AD pathology assessed through CSF tau and neurofilament light (NfL) [22, 23]. These associations imply that a subtle decline in motor functions may serve as an early proxy for accumulating AD neuropathology, reinforcing the value of sensorimotor assessments in preclinical dementia detection.

We hypothesize that poorer physical function, reflected in slower gait speed, weaker grip strength, and longer reaction time, will be associated with greater Alzheimer’s disease–related biomarker burden, including lower CSF beta-amyloid 42/40 ratio (Aβ_42/40), higher CSF phosphorylated tau to beta-amyloida 42 ratio (pTau/Aβ₄₂) and total tau to beta-amyloid 42 ratio (tTau/Aβ₄₂) and elevated plasma NfL_42/40, at baseline. Furthermore, we hypothesize that these sensorimotor impairments will correspond to lower baseline cognitive function and predict a faster rate of cognitive decline over time.

Methods

Participants and cohort

This prospective longitudinal observational study included community-dwelling older adults who were cognitively normal at baseline (CDR = 0) and did not progress to dementia using the Clinical Dementia Rating (CDR > 0) during follow-up. Participants are enrolled in The DRIVES Project at Washington University in St. Louis (R01AG056466, R01AG068183, R01AG067428). Eligible participants were aged 65 years or older and native English speakers. All participants completed a comprehensive neuropsychiatric assessment and physical function measurement evaluation at baseline. Annual follow-up assessments of cognitive function were conducted. Participants provided written informed consent before enrollment. This study was approved by the Washington University Institutional Review Board and the Human Research Protection Office (IRB #202010214 and #202003209).

Demographic data, including race and education level, were self-reported by participants. Socioeconomic disadvantage was quantified using the Area Deprivation Index (ADI) national ranking, which captures neighborhood-level socioeconomic conditions. Participants’ home addresses were geocoded at the census block-group level to derive ADI scores. The ADI incorporates factors such as income, education, employment, and housing quality, providing a comprehensive measure of community-level disadvantage.

Cognitive assessment and APOE genotyping

Cognitive function was assessed using the Preclinical Alzheimer’s Cognitive Composite (PACC), derived from four neuropsychological assessments: Animal Fluency [24], the Selective Reminding Test – Free Recall [25], and Trail Making Tests A and B [26]. Scores from the Trail Making Tests were reversed, with higher values indicating lower cognitive function. These tests evaluate verbal fluency, episodic memory, attention, processing speed, spatial recognition, and executive function. Raw scores from each test were standardized into z-scores based on the baseline sample mean and standard deviation from the present cohort. Individual z-scores were then averaged (unweighted) to compute the overall PACC score, with higher scores indicating better cognitive performance. This approach, consistent with prior applications of the PACC in community-based and preclinical Alzheimer’s disease studies, allows for sensitive detection of early cognitive variability among cognitively normal older adults [5, 27, 28]. APOE ε2, ε3, and ε4 alleles were determined using SNP genotyping conducted at the Department of Neurology, Washington University, with standard quality control protocols.

Physical function assessments

Grip strength was measured using a Jamar hydraulic hand dynamometer. Participants were seated with their elbows flexed at 90 degrees without arm support. Grip strength was recorded three times for each hand, alternating sides, resulting in a total of six measurements. The average of these measurements was used for analysis. If only one hand was available due to injury, prosthesis, or arthritis, measurements from that single hand were used. Gait speed was evaluated using a 10-m walk test [29]. Participants were instructed to walk at their usual pace along a clearly marked 10-m path. Timing was recorded using a stopwatch for three consecutive trials, and the average speed (in meters per second) was calculated. Participants were permitted to use walking aids (e.g., walking sticks or canes) if commonly used. Reaction time was assessed using the Deary-Liewald computerized reaction time test [30]. Participants completed a practice session of eight trials, with instructions emphasizing rapid response to an "X" appearing on a computer screen by pressing any key on the keyboard. After satisfactory completion of the practice trials, participants proceeded to the formal test, which consisted of 20 trials. Reaction times were recorded in milliseconds, and the median reaction time was used for analysis to minimize skewness. Body mass index (BMI) was calculated from height and weight measurements (kg/m²). Participants were weighed on a calibrated scale, and height was measured standing upright. Participants removed heavy clothing but retained shoes during measurement.

CSF and plasma biomarkers

Cerebrospinal fluid (CSF) samples were collected via lumbar puncture by experienced neurologists following an overnight fast. Approximately 20–35 mL of CSF was obtained around 8:00 am using either a 22-gauge Sprotte spinal needle or a 22-gauge Quincke-type needle. Samples were immediately inverted, centrifuged, aliquoted, and stored at −80 °C. Biomarkers measured included Aβ_42/40, tTau/Aβ₄₂, and pTau/Aβ₄₂, all analyzed using automated electrochemiluminescence immunoassays on the Lumipulse G1200 platform (Fujirebio, Tokyo, Japan). CSF biomarkers were dichotomized based on previously validated thresholds to classify participants as biomarker-positive or -negative. Positivity was defined as Aβ_42/40 ratio < 0.0673, pTau/Aβ₄₂ ratio > 0.0649, and tTau/Aβ₄₂ ratio > 0.488 [5, 31]. Plasma neurofilament light chain (NfL) concentrations were measured using Quanterix Nf-Light assay kits on an HD-X analyzer. Biomarker data used for analysis included samples collected within two years of the baseline assessment.

Statistical analysis

Descriptive statistics were calculated to characterize baseline demographics, clinical features, cognitive function, and biomarker distributions. Participants were categorized based on PACC scores, with "low cognitive function" defined as scores at or more than 0.5 standard deviations below the baseline sample mean, and "high cognitive function" defined as scores above this threshold. Between-group differences were examined using t-tests or chi-square tests as appropriate. We analyzed cross-sectional associations using ordinary least squares general linear models with complete-case analysis (listwise deletion). Baseline cognition (PACC) and each biomarker outcome (plasma NfL, CSF pTau/Aβ₄₂, tTau/Aβ₄₂, and Aβ_42/40) were modeled with the same predictor set: simple reaction time (SRT), 10-m walk time (as the gait-speed measure), grip strength, age at assessment, sex, years of education, Area Deprivation Index (national rank), race, body mass index (BMI), and APOE ε4 carrier status. Two-sided tests with α = 0.05 were used. Statistical assumptions, including normality and multicollinearity, were verified using residual diagnostics and variance inflation factor (VIF) analyses.

Longitudinal analyses were conducted using linear mixed-effects models (LME) to examine whether baseline sensorimotor performance predicted cognitive change over approximately four years of follow-up. Baseline gait speed was dichotomized to slow gait and normal gait, with a cut-off speed of 0.8 m/s for longitudinal analysis [7, 32, 33]. A cutoff of 0.5 standard deviations above the sample mean from the baseline subset was used for the median simple reaction time test, with values exceeding this threshold classified as slow reaction time for baseline data classification for longitudinal analysis. Baseline grip strength was categorized using sex-specific cutoffs established by the Foundation for the National Institutes of Health (FNIH) Sarcopenia Project, with low grip strength defined as < 26 kg for men and < 16 kg for women [34]. Models included random intercepts to account for repeated measures within participants and were adjusted for the same covariates as those in the cross-sectional analyses. Statistical significance was set at a p-value less than 0.05. All analyses were performed using available data without imputation for missing values.

Results

Participant characteristics

The analytic sample included 246 community-dwelling older adults (mean age = 74.9 years, SD = 5.17) enrolled at baseline. The majority identified as non-Hispanic White (86.6%), with 13.4% identifying as Black. Education averaged 16.4 years (SD = 2.37). Approximately half of the sample consisted of females (48.8%). APOE genotyping revealed that 31.3% of participants carried at least one ε4 allele. Participants were classified into two groups based on their PACC score, using a threshold of 0.5 standard deviations below the mean: 191 participants were categorized as having high cognitive function, and 55 were categorized as having low cognitive function.

Participants with low PACC scores were slightly older (mean = 76.2 vs. 74.5 years, p = 0.049) and had a higher proportion of Black individuals (25.5% vs. 9.9%, p = 0.006). Participants in the low-PACC group demonstrated significantly slower gait speed on the 10-m walk test (mean = 9.28 s vs. 8.31 s, p = 0.008). Among the 246 participants, 61 (24.8%) were CSF Aβ_42/40-positive, 52 (21.1%) were pTau/Aβ₄₂-positive, and 51 (20.7%) were tTau/Aβ₄₂-positive. Plasma NfL values were only available for 171 participants, with no significant differences between the groups (see Table 1).

Table 1.

Baseline characteristics by cognitive performance group

	Total (N = 246)	Normal PACCa (N = 191)	Low PACC (N = 55)	P-value
Age (Mean (SD))	74.9 (5.17)	74.5 (4.94)	76.2 (5.75)	.049*
Sex
Female	120 (48.8%)	95 (49.7%)	25 (45.5%)	.684
Race
non-Hispanic Whites	213 (86.6%)	172 (90.1%)	41 (74.5%)	.006**
African American	33 (13.4%)	19 (9.9%)	14 (25.5%)
Education (Mean (SD))	16.4 (2.37)	16.4 (2.36)	16.1 (2.41)	.436
APOE ε4b (Positivity)
Case	77 (31.3%)	62 (32.5%)	15 (27.3%)	.708
BMIb Mean (SD)	27.9 (4.94)	28.0 (4.96)	27.8 (4.92)	.837
ADIb (Mean (SD))	42.6 (24.3)	41.4 (23.5)	46.6 (26.8)	.192
Grip Strength (Mean (SD))	58.8 (21.2)	59.2 (20.8)	57.4 (22.3)	.004**
Gait speed (10 m Walk Test) (Mean (SD))	8.53 (1.87)	8.31 (1.59)	9.28 (2.49)	.008**
Simple Reaction Time (Mean (SD))	347 (85.8)	345 (82.7)	353 (96.6)	.558
Plasma NfLb (Mean (SD))	12.1 (9.05)	11.9 (8.74)	12.8 (10.2)	.659
CSF pTaub/Aβ42b Ratio
Positive	52 (21.1%)	37 (19.4%)	15 (27.3%)	.079
CSF tTaub/Aβ42 Ratio	62.8 (39.6)	49.3 (29.2)	90.7 (43.3)
Positive	51 (20.7%)	37 (19.4%)	14 (25.5%)	.155
CSF Aβ42/40b Ratio
Positive	61 (24.8%)	46 (24.1%)	15 (27.3%)	.339

a. PACC groups were defined using a threshold of 0.5 SD below the baseline sample mean

b. PACC = Preclinical Alzheimer Cognitive Composite; APOE = Apolipoprotein E; BMI = Body mass index (kg/m²); ADI = Area Deprivation Index (national percentile); NfL = Neurofilament light chain; Aβ₄₂ = Amyloid-beta 42; Aβ₄₀ = Amyloid-beta 40; pTau = phosphorylated tau; tTau = total tau

c. p-values reflect t-tests or chi-square tests comparing low vs. normal PACC groups

*p <.05, **p <.01, ***p <.001

Cross-sectional general linear model results

In the baseline multivariable general linear model examining cognitive performance (PACC score), older age (β = −0.021, p = 0.015), male sex (β = −0.293, p = 0.010), and slower simple reaction time (β = −0.0015, p = 0.028) were significantly associated with lower PACC scores. Educational attainment, BMI, ADI rank, gait speed, grip strength, race, and APOE status were not significant predictors of the outcome.

In models assessing plasma and CSF biomarkers as dependent variables, higher plasma NfL concentrations were associated with higher ADI deprivation rank (β = −0.073, p = 0.028), lower BMI (β = −0.406, p = 0.012), slower gait speed (β = 1.603, p = 0.025), and APOE ε4 carrier status (β = 3.062, p = 0.049). Higher CSF pTau/Aβ₄₂ ratio was associated with slower gait speed (β = 0.0108, p = 0.032) and APOE ε4 carrier status (β = 0.0382, p = 0.001). Higher CSF tTau/Aβ₄₂ ratio was associated with lower BMI (β = −0.0158, p = 0.028), slower gait speed (β = 0.082, p = 0.008), and APOE ε4 carrier status (β = 0.251, p < 0.001). Lower CSF Aβ_42/40 ratio was associated with older age (β = −0.00087, p = 0.027) and APOE ε4 carrier status (β = −0.024, p < 0.001). (See Table 2).

Table 2.

Associations between demographic factors and sensorimotor function measures with biomarkers and PACC scores

	Pacc score (n = 183)	Plasma NfL (n = 141)	CSF pTau/Aβ42 Ratio (n = 128)	CSF tTau/Aβ42 Ratio (n = 128)	CSF Aβ42/40 Ratio (n = 128)
Age	−0.02* [−0.04, −0.004]	0.23 [−0.08, 0.54]	0.001 [−0.001, 0.004]	0.004 [−0.01, 0.02]	−0.001* [−0.002, −0.0001]
Sex (Male)	−0.29* [−0.51, −0.07]	0.68 [−3.35, 4.70]	−0.005 [−0.03, 0.02]	0.01 [−0.18, 0.19]	−0.001 [−0.01, 0.008]
Education (Years)	0.01 [−0.02, 0.05]	−0.15 [−0.81, 0.51]	−0.002 [−0.006, 0.003]	−0.01 [−0.04, 0.02]	0.0005 [−0.001, 0.002]
Race (nHW)	0.25 [−0.02, 0.53]	0.84 [−4.35, 6.02]	0.02 [−0.02, 0.06]	0.14 [−0.12, 0.40]	−0.006 [−0.02, 0.008]
APOE ε4 (Case)	0.03 [−0.14, 0.19]	3.06* [0.02, 6.12]	0.04** [0.01, 0.06]	0.25*** [0.11, 0.39]	−0.02 [−0.03, −0.02]
BMI	0.004 [−0.01, 0.02]	−0.41* [−0.72, −0.09]	−0.001 [−0.003, −0.0004]	−0.01* [−0.03, 0.01]	0.0002 [−0.0005, 0.0009]
ADI	−0.002 [−0.01, 0.0016]	−0.07* [−0.14, −0.01]	0.0001 [−0.003, 0.0007]	0.0004 [−0.002, 0.003]	0.00001 [−0.0002, 0.0002]
Simple Reaction Time	−0.0015* [−0.003, −0.0002]	0.03 [−0.004, 0.06]	−0.00005 [−0.0003, 0.0002]	−0.0003 [−0.002, 0.001]	0.00003 [−0.00004, 0.0001]
Grip Strength	0.003 [−0.01, 0.01]	0.06 [−0.04, 0.16]	0.0005 [−0.0003, 0.0013]	0.003 [−0.001, 0.008]	0.00002 [−0.0002, 0.0003]
Gait speed (10 m Walk Test)	−0.04 [−0.10, 0.03]	1.60* [0.21, 2.99]	0.011* [0.001, 0.021]	0.08** [0.02, 0.14]	−0.0003 [−0.004.0.003]

PACC = Preclinical Alzheimer Cognitive Composite; NfL = Neurofilament light chain; Aβ₄₂ = Amyloid-beta 42; Aβ₄₀ = Amyloid-beta 40; pTau = phosphorylated tau; tTau = total tau; APOE = Apolipoprotein E; BMI = Body mass index (kg/m²); ADI = Area Deprivation Index (national percentile); Estimates reflect β coefficients and 95% confidence intervals

^*p <.05, **p <.01, ***p <.001

Longitudinal analyses of cognitive trajectories

Among the 192 participants with at least one follow-up cognitive assessment, longitudinal cognitive trajectories were modeled over a mean follow-up duration of 4.01 years using linear mixed-effects models. Separate models were estimated for each sensorimotor variable.

Participants with normal grip strength had significantly higher PACC scores over time (β = 0.268, p = 0.002). However, the interaction between grip strength and time was not statistically significant (β = −0.004, p = 0.653). In the second LME model, normal gait speed at baseline was significantly associated with higher PACC scores (β = 0.630, p < 0.001) and a slower rate of cognitive decline over time (interaction β = 0.069, p = 0.008). These findings suggest both level and slope effects of gait speed on cognition over time (see Fig. 1). Slower reaction time, defined as greater than 0.5 SD above the baseline sample median, was associated with lower PACC scores longitudinally (β = −0.002, p < 0.001). However, there was no significant interaction between reaction time and time (p = 0.28). Across all longitudinal models, age, sex, and race remained consistent covariates associated with cognitive function. APOE genotype and BMI were not significantly associated with cognitive change (see Table 3).

Fig. 1.

Predicted PACC trajectories are plotted by gait speed classification, with those with gait speed below 1 m/second as “Low gait speed” and those above this threshold as “Normal Gait Speed”. Participants with normal gait speed had a higher starting cognitive score (intercept = 1.255) compared to those with low gait speed (intercept = 0.880). Over time, individuals with low gait speed demonstrated significantly faster cognitive decline (slope =  − 0.0426) than those with normal gait speed (slope =  − 0.0116). Estimates are derived from a linear mixed effects model adjusted for age, sex, education, race, ADI, BMI, APOE, and random intercepts for participants. Line graph showing predicted PACC score trajectories over time for two gait speed groups. The blue line represents participants with normal gait speed (≥ 1 m/second), who initially have higher cognitive scores and exhibit a slower rate of decline. The red line represents participants with low gait speed (< 1 m/second), who started with lower cognitive scores and declined more rapidly over the follow-up period

Table 3.

Association between baseline sensorimotor function and longitudinal PACC score

	Grip Strength	Gait speed (10 m Walk Test)	Simple Reaction Time
Age	−0.02*** [−0.04, −0.01]	−0.02** [−0.03, −0.01]	−0.03*** [−0.04, −0.02]
Sex (Male)	−0.20** [−0.33, −0.06]	−0.21** [−0.34, −0.08]	−0.17* [−0.31, −0.02]
Education (Yrs)	0.03* [0.00, 0.06]	0.02 [0.00, 0.05]	0.02 [−0.01, 0.05]
Race (nHW)	0.49*** [0.28, 0.69]	0.41*** [0.20, 0.62]	0.32** [0.09, 0.55]
APOE-ε4 (Case)	0.004 [−0.13, 0.14]	0.008 [−0.05, −0.128]	0.01 [−0.13, 0.16]
BMI	−0.003 [−0.01, 0.01]	−0.003 [−0.013, 0.006]	−0.01 [−0.02, 0.003]
ADI	−0.001 [−0.004, 0.002]	−0.001 [−0.004, 0.001]	−0.002 [−0.01, 0.001]
Grip Strength
Low grip strength	−0.27** [−0.43, −0.10]
10 m Walk Test
Slow gait speed		−0.38*** [−0.59, −0.16]
Simple Reaction Time			−0.001* [−0.003, −0.0004]
Slow reaction time
Interaction with time	0.004 [−0.01, 0.02]	−0.03* [−0.06, −0.006]	−0.00006 [−0.0003, 0.0001]

PACC = Preclinical Alzheimer Cognitive Composite; APOE = Apolipoprotein E; BMI = Body mass index (kg/m²); ADI = Area Deprivation Index (national percentile); nHW = non-Hispanic White

Models are adjusted for age, sex, education, BMI, race, ADI, and APOE-ε4 status

Low grip strength: < 26 kg (men), < 16 kg (women); Slow gait speed: < 0.8 m/s; Slow reaction time: > 0.5 SD above baseline sample median

Estimates reflect β coefficients and 95% confidence intervals

^*p <.05, **p <.01, ***p <.001

Discussion

In this prospective study of community-dwelling older adults, stronger grip strength, faster gait speed, and quicker reaction time were associated with better cognitive performance cross-sectionally and longitudinally. Unlike prior studies, we examined multiple sensorimotor domains simultaneously in a cognitively normal cohort and linked them with both longitudinal cognition and AD-related biomarkers. In our multivariable models, slower reaction time was significantly associated with lower baseline cognitive composite scores, consistent with prior evidence that psychomotor slowing reflects early declines in neural processing efficiency linked to cognitive deterioration [15, 35]. Reaction time tasks, which measure basic cognitive speed, have been shown to identify subtle cognitive changes before clear clinical symptoms appear, highlighting their potential as a screening tool in primary care or community settings [10, 16, 36]. In addition, older age and male sex were independently associated with lower cognitive scores, consistent with established demographic risk factors for cognitive impairment. Gait speed showed significant associations with CSF tau and p-tau burden, indicating its potential to capture underlying AD-related neurodegenerative processes. Prior studies have linked slower gait to tau and amyloid pathology, but most have focused on individuals with cognitive impairment [15, 16]. Our findings extend the existing literature by demonstrating these associations in a cognitively normal, community-based cohort, utilizing a panel of CSF/plasma biomarkers and adjusting for relevant covariates. These findings support the hypothesis that gait speed reflects shared neurodegenerative processes underlying both motor and cognitive decline. Slower gait may indicate early white matter degeneration, cortical atrophy, and tau-related pathology within frontal–subcortical and parietal networks critical for mobility and executive function. The relationship is likely bidirectional, as emerging evidence from longitudinal neuroimaging studies suggests that early neurodegenerative changes can simultaneously disrupt motor and cognitive systems [21, 22]. Together, these results reinforce the biological plausibility of gait speed as a sensitive, non-invasive marker of preclinical Alzheimer’s disease pathology and early brain vulnerability [37].

Of the sensorimotor measures examined longitudinally, slow baseline gait speed was the only impairment that predicted both lower average cognitive scores across follow-up and a faster rate of decline. In contrast, low grip strength and slow reaction time were associated only with lower cognitive function across follow-up. Participants with normal gait speeds had higher PACC scores at baseline and experienced slower cognitive decline. These longitudinal results complement existing literature establishing gait speed as a robust marker of cognitive health [20, 32]. Indeed, gait speed has been described as a sensitive integrative measure reflecting the integrity of multisystem functions, including motor control, sensory integration, and higher-level cognitive processes [7, 8]. Based on our longitudinal models, participants with slow gait speed (< 0.8 m/s) had, on average, 0.38 lower PACC scores than those with normal gait speed, corresponding to roughly a 0.4 SD difference in cognitive performance. Similarly, those with low grip strength scored 0.27 points lower on the PACC compared with participants with normal grip strength. These differences represent subtle yet functionally relevant disparities in cognition among otherwise cognitively normal older adults.

Clinically, these findings underscore the value of incorporating simple physical assessments, such as gait speed, grip strength, and reaction time, into routine primary care and geriatric screenings to help identify older adults at elevated risk for cognitive impairment. Gait speed thresholds of less than 0.8 m/s are widely recognized as indicators of slow gait and have been associated with increased risk of frailty, disability, and cognitive decline [33]. Our longitudinal models employed this threshold, consistent with prior meta-analyses in geriatric and dementia research, which supports its validity in identifying older adults at elevated risk for cognitive decline. Individuals walking below this cut-off may warrant closer monitoring or early preventive intervention. These measures are already routinely evaluated by allied health professionals, particularly physical and occupational therapists, who are uniquely positioned to detect early signs of functional decline. Primary care providers are encouraged to evaluate older adults for falls and can incorporate gait measures into their annual medicare wellness visits [38]. Their expertise in assessing mobility, strength, and sensorimotor integration supports early identification of cognitive risk, enabling timely, non-pharmacologic interventions to promote healthy aging and prolong independence [39, 40]. Our biomarker findings suggest that cognitively normal older adults exhibiting biomarker positivity may already be experiencing subtle cognitive alterations, which warrants more comprehensive cognitive testing, even in the absence of clinical symptoms. In this context, emerging tools such as the Ambulatory Research in Cognition platform and naturalistic driving behavior studies from the DRIVES Project offer valuable paradigms for assessing real-world cognitive functioning beyond traditional clinic-based measures [41, 42]. These approaches may improve early detection of preclinical Alzheimer’s disease by capturing subtle, everyday cognitive lapses that precede measurable decline on standardized tests. Identifying these individuals may facilitate the development of targeted preventive strategies, including lifestyle interventions, physical activity programs, or clinical trials involving novel therapeutics aimed at the preclinical stages of cognitive decline.

This study has notable strengths, including its longitudinal design, detailed sensorimotor assessments, comprehensive cognitive evaluations, and integration of robust biomarker analyses. However, limitations should also be acknowledged. The biomarker data, although insightful, were limited by missing data, which restricted their inclusion in longitudinal analyses and potentially biased the cross-sectional findings. Furthermore, our cohort was predominantly non-Hispanic White and well-educated, limiting the generalizability of the findings to more diverse populations. Our primary outcome, the PACC, contains constructs that are conceptually similar to some of our predictor variables; however, these measures capture distinct domains, sensorimotor performance versus cognitive function, and are obtained through different modalities, reducing the likelihood of redundancy and eliminating the need for statistical adjustment. Comorbidities, which may confound associations between sensorimotor function and cognition, were not included in our analyses. Finally, while observational designs elucidate important associations, they do not permit causal inference. Future work will aim to replicate these findings in larger and more diverse cohorts to assess generalizability and to extend follow-up to examine transitions to mild cognitive impairment or dementia.

Our findings demonstrate that baseline physical function measures, particularly grip strength, gait speed, and reaction time, are predictors of cognitive function among cognitively normal older adults. These functional markers, complemented by neurodegenerative and amyloid biomarkers, may enhance early detection strategies for cognitive decline, offering opportunities for timely preventive interventions. Integrating these simple, accessible assessments into routine geriatric care could substantially improve cognitive health surveillance and dementia prevention efforts among aging populations.

Author contribution

Semere Bekena: Conceptualization, data curation, formal analysis, visualization, writing original draft, and review and editing.

Ramkrishna K. Singh: Methodology, data interpretation, writing – review and editing.

Yiqi Zhu: Formal analysis, statistical validation, writing – review and editing.

David B. Carr: Methodology, writing – review and editing.

Ganesh M. Babulal: Conceptualization, methodology, validation, supervision, funding acquisition, writing – review and editing.

Funding

This work was supported by grants awarded to Ganesh M. Babulal from the National Institutes of Health (NIH) and the National Institute on Aging (NIA) (R01 AG068183, R01 AG067428, R01 AG074302). The funding agencies had no role in the design, conduct, data collection, management, analysis, or interpretation of the study, nor in the preparation, review, or approval of the manuscript.

Data availability

The data supporting this study's findings are available from the corresponding author upon reasonable request.

Declarations

Ethics and consent to participate

Written informed consent was obtained from all participants prior to enrollment in the DRIVES Project. All study procedures were approved by the Washington University Institutional Review Board and the Human Research Protection Office (IRB #202010214 and #202003209).

Competing interest

• Semere Bekena, MD, MPH: No conflicts to disclose

• Ramkrishna K. Singh, MBBS, MPH: No conflicts to disclose

• Yiqi Zhu, PhD: No conflicts to disclose

• David B. Carr, MD: No conflicts to disclose

• Ganesh M. Babulal, PhD, OTD: No conflicts to disclose