Associations between vestibular perception and cognitive performance in healthy adults

Abstract

Objectives:

A growing body of evidence has linked vestibular function to the higher-order cognitive ability in aging individuals. Past evidence has suggested unique links between vestibular function and cognition on the basis of end-organ involvement (i.e., otoliths vs. canals). However, past studies have only assessed vestibular reflexes despite the diversity of vestibular pathways. Thus, this exploratory study aimed to assess associations between vestibular perception and cognition in aging adults to determine potential relationships.

Design:

Fifty adults (21–84 years; mean=52.9, SD=19.8) were included in this cross-sectional study. All participants completed a vestibular perceptual threshold test battery designed to target perception predominantly mediated by each end-organ pair and intra-vestibular integration: 1 Hz y-translation (utricle), 1 Hz z-translation (saccule), 2 Hz yaw rotation (horizontal canals), 2 Hz right anterior, left posterior (RALP), and left anterior, right posterior (LARP) tilts (vertical canals), and 0.5 Hz roll tilt (canal-otolith integration). Participants also completed standard assessments of cognition and path integration: Digit Symbol Substitution Test (DSST), Trail Making Test (TMT), and the Gait Disorientation Test (GDT). Associations were assessed using Spearman rank correlation, and multivariable regression analyses.

Results:

For correlation analyses, DSST correlated to RALP/LARP tilt, roll tilt, and z-translation. TMT-A only correlated to z-translation, and TMT-B correlated to roll tilt and z-translation after correcting for multiple comparisons. GDT correlated to RALP/LARP tilt and y-translation. In age-adjusted regression analyses, DSST and TMT-B were associated with z-translation thresholds and GDT was associated with y-translation thresholds.

Conclusions:

In this cross-sectional study, we identified associations between vestibular perceptual thresholds with otolith contributions and standard measures of cognition. These results are in line with past results suggesting unique associations between otolith function and cognitive performance.

1. Introduction

The vestibular system consists of the semicircular canals (SCC), which encode rotation, and the otolith organs which encode translation and tilt (i.e., re-orientation relative to gravity) of the head (Goldberg, 2012). Vestibular function has been traditionally viewed as integral to several fundamental responses, including gaze stability, self-motion perception, autonomic regulation, and balance performance (Wolfe et al., 2006). This diversity of function is partially reflected by widespread projections to numerous subcortical and cortical areas (Hitier et al., 2014). A growing body of evidence also has suggested that degradations in vestibular function are linked to higher cortical processing, particularly to cognitive function (for reviews see (Agrawal, Smith, et al., 2019; Bosmans et al., 2021; Dobbels, Peetermans, et al., 2019; L. J. Smith et al., 2024)

Associations between vestibular function and cognitive performance have been identified in healthy aging individuals (i.e., in the absence of pathologic vestibular loss or cognitive impairment) suggesting that age-related variability in cognitive function may be related to age-related vestibular loss. This includes population-based studies, such as the National Health and Nutrition Examination Survey (NHANES) and Baltimore Longitudinal Study of Aging (BLSA). In community dwelling adults assessed in the NHANES cohort, balance performance with predominant vestibular sensory contributions (i.e., eyes closed, standing on foam) was related to decreased performance on the Digit Symbol Substitution Test (DSST) (Semenov et al., 2016). Similarly, in the BLSA cohort, cervical vestibular evoked myogenic potential (cVEMP) responses, reflecting integrity of the saccule and the vestibulo-colic reflex (VCR), were associated with decreased performance on the Trail Making Test (TMT) (Bigelow et al., 2015).

Further cross-sectional studies suggest age-related vestibular declines, as quantified by reduced vestibular reflexive function including cVEMPs and the video head impulse test (vHIT), are associated with age-related impairments in spatial navigational abilities in older adults (Biju et al., 2021; Xie et al., 2017) potentially reflecting links from the vestibular system to the hippocampus (Brandt et al., 2005; Cohen et al., 2022; P. F. Smith et al., 2015). This includes age-related changes in path integration, a component of spatial navigation, which reflects the continuous updating of distance and direction traveled using pertinent vestibular, visual, and proprioceptive sensory cues (Adamo et al., 2012; Anson et al., 2019; Xie et al., 2017). These proposed links between vestibular function and spatial cognition are further bolstered by substantial preclinical work in animal models demonstrating long-term deficits in spatial memory secondary to chemical or surgical bilateral vestibular ablation (P. F. Smith, 2017; P. F. Smith & Zheng, 2013) which persist despite compensation of the vestibulo-spinal reflex (VSR) and/or vestibulo-ocular reflex (VOR) (Baek et al., 2010; P. F. Smith et al., 2015; Zheng et al., 2007, 2009).

Additionally, a link between vestibular function and cognition is supported by results in patients with vestibular loss and cognitive dysfunction. In adults with bilateral vestibular loss (BVL), consistent deficits in spatial cognition have been identified (Dobbels, Peetermans, et al., 2019). While less consistent, changes in non-spatial cognition dependent on patient population have been reported. Additional cognitive processes have also shown changes in BVL patients including executive function, immediate memory, and processing speed (Bigelow & Agrawal, 2015; Brandt et al., 2005; Dobbels, Peetermans, et al., 2019; Schöne et al., 2024). As well, in adults with mild cognitive impairment (MCI) and Alzheimer’s disease (AD), a higher prevalence of vestibular loss has been identified in comparison to older adults with preserved cognitive function (Bosmans et al., 2021; Cohen et al., 2022; Guo et al., 2024; Harun et al., 2016; Hung et al., 2024; Wei et al., 2018).

However, all past studies linking vestibular function to higher-order cognitive function have focused on reflexive vestibular pathways, particularly the VCR and VOR, or balance performance as a metrics of vestibular function and have not rigorously assessed potential links to other vestibular pathways. Vestibular reflexive and higher-order (i.e., those with cortical projections via the thalamus) pathways are anatomically distinct and exhibit distinct behaviors (Haburcakova et al., 2012; Merfeld, 2005; Merfeld et al., 2005). Vestibular-higher order assessments, including measures of vestibular perceptual sensitivity, identify the ability to use sensory information to generate a percept of self-motion/orientation(Cullen & Chacron, 2023). As both cognition and higher-order vestibular behavior include cortical contributions, they may show unique associations not captured in vestibular reflex-based assays.

Additionally, there is also a growing body of growing body of evidence suggesting a unique role of otolith function in higher-order cognitive performance (for review see Smith 2019), which has not been extensively investigated in healthy aging adults. A recent systematic review of adults with MCI or AD identified that cognitive loss was predominantly associated with reflexive measures quantifying loss of otolith function (i.e., cVEMP, ocular VEMP (oVEMP)), whereas semi-circular canal function was more-so preserved (Agrawal et al., 2020; Bosmans et al., 2021). However, other studies of patients with bilateral vestibular loss (BVL) have shown links between cognition and both decreases in otolith and semi-circular canal function (Pineault et al., 2020), potentially reflecting the diversity in peripheral end-organ involvement in these individuals.

Thus, this study aimed to assess potential associations between vestibular perception and cognition in healthy adults. We hypothesized that age-related changes in vestibular perception would be identified that would correlate to cognitive measures. The secondary aim of this study was to explore modality specific (i.e., measures predominantly reflecting canal or otolith function) associations between vestibular perception and cognition. We hypothesized that the largest associations would be seen for perceptual measures reflecting perception predominantly mediated by the otoliths (i.e., utricle and saccule). Since our cognitive test battery was partially selected based on past population based studies which identified an association between reflexive vestibular measures and cognition in healthy aging (Bigelow et al., 2015; Semenov et al., 2016), we also aimed to explore which cognitive measures, reflecting different domains of cognitive performance (e.g., spatial navigation, processing speed), displayed preferential associations with vestibular perception.

2. Methods

2.1. Participants

A total of 51 participants were recruited to participate in this study. In order to ensure inclusion of a wide age range to capture age related changes in vestibular perception and cognitive performance, recruitment was targeted to three groups: younger adults (18–39 years of age), middle-aged adults (40–64 years of age), and older adults (65 to 89 years of age). The delineation of middle-aged adults was made based off of documented changes in vestibular perception after the approximate age of 40 (Bermúdez Rey et al., 2016). The demarcation of older adults as participants ≥ 65 years of age was made based upon the definition adopted by the American Medical Association (Lundebjerg et al., 2017). One older adult participant (78-year-old female) was identified as an outlier and excluded from analyses as vestibular thresholds were ~2–20x higher than the mean and regression diagnostics indicated high influence. Demographic information from the remaining 50 participants is provided in Table 1.

Table 1.

Participant demographics and descriptive statistics (mean ± 1 SD) are shown for the primary variables of interest. RALP/LARP tilt thresholds represent averages of both thresholds from participants (n=44) which were able to complete. One middle aged and five older adults were unable to complete RALP/LARP tilt thresholds as the psychophysical staircase exceeded the displacement limits of the motion device. GDT=Gait Disorientation Test; DSST=Digit Symbol Substitution Test; SAGE=Self-administered Gero-cognitive Exam; LARP =left anterior, right posterior; RALP=right anterior, left posterior; TMT-A=Trail Making Test Part-A; TMT-B=Trail Making Test Part-B.

	Young Adult (18–39 years)	Middle Aged (40–64 years)	Older Adult (65–85 years)	Total
N (Female)	17(9F)	15 (11F)	18(11F)	50(31F)
Age	29.7 ± 5.4	52.1 ± 6.0	74.5 ± 5.6	52.9 ± 19.8
Education
High School	0(0F)	1(0F)	2(0F)	3(0F)
College Degree	12(6F)	6(6F)	6(5F)	25(13F)
Graduate Degree	5(2F)	8(5F)	10(6F)	22(17F)
SAGE	21.59 ± 0.80	21.13 ± 1.46	20.53 ± 1.81	21.06 ± 1.48
DSST	74.35 ± 9.17	63.93 ± 12.84	42.72 ± 10.87	59.84 ± 17.36
TMT-A	19.83 ± 5.39	21.19 ± 6.18	34.48 ± 13.48	25.53 ± 11.39
TMT-B	40.36 ± 8.86	48.79 ± 17.22	78.87 ± 23.96	56.76 ± 24.53
GDT	0.57 ± 0.9	1.39 ± 1.47	3.20 ± 2.82	1.78 ± 2.25
Yaw rotation (deg/s)	0.717± 0.356	0.802 ± 0.532	1.475 ± 1.067	1.026 ± 0.803
RALP/LARP tilt (deg/s)	0.611 ± 0.230	0.933 ± 0.294	1.876 ± 1.194	1.069 ± 0.835
Roll tilt (deg/s)	0.815 ± 0.271	1.053 ± 0.537	1.646 ± 1.001	1.188 ± 0.775
Y-translation (cm/s)	0.665 ± 0.402	0.775 ± 0.267	1.658 ± 1.207	1.056 ± 0.899
Z-translation (cm/s)	1.593 ± 0.884	2.812 ± 1.515	5.661 ± 2.962	3.391 ± 2.649

Prior to participation, all participants completed questionnaires assessing medical history and general health background. Participants were excluded on the basis of report of vestibular disorder (excluding resolved BPPV greater than 6 months prior), neurological disorder (e.g., Parkinson’s disease), major health condition (e.g., cancer, recent surgery), uncontrolled cardiovascular disease, uncorrected visual impairment, and/or recent (<6 months) orthopedic injury. Additionally, participants were excluded if they exceeded the weight limit (>250 lbs.) of our motion platform. All participants provided written informed consent prior to participation in laboratory procedures and participants were compensated for their time. All study procedures were approved by the Institutional Review Board.

2.2. Vestibular Thresholds

Participants completed a vestibular threshold test battery. General methodology for vestibular thresholds has been previously described in detail (Chaudhuri & Merfeld, 2013; Grabherr et al., 2008; Karmali et al., 2016) and has been used extensively by our lab (Karmali et al., 2017; Kobel, Wagner, & Merfeld, 2021; Wagner, Kobel, & Merfeld, 2022; Wagner, Kobel, Tajino, et al., 2022) and other investigators (MacNeilage et al., 2010; Roditi & Crane, 2012; Suri & Clark, 2020). In brief, perceptual thresholds were quantified using a one-interval direction recognition task (e.g., “did I turn right or left?”). For each threshold measure, described below, 100 trials or 100 motions were completed which, depending on frequency and type of motion, took ~8–16 minutes each. Participants were provided a break after ~30 minutes of testing or sooner if requested.

Participants were restrained using a four-point harness and motorcycle helmet which was rigidly attached to the motion platform. To maximize vestibular contributions to motion perception, testing was completed in a light-tight room to remove visual cues. Auditory contributions to motion perception were also reduced via presentation of ~60 dB sound pressure level (SPL) of auditory white noise presented during motion stimuli. Additional passive sound attenuation (~20 dB SPL) was provided via use of insert earphones.

Participants were seated in a chair mounted to a MOOG (6DOF2003E; East Aurora, New York) six degree of freedom platform. Motion stimuli were delivered via the MOOG platform. Stimuli were single cycles of sinusoidal acceleration ([a(t)=A sin(2πft)], A=amplitude, f=motion frequency) in which peak acceleration (A), peak velocity (v_p), and total displacement are proportional (v_p=A/πf= 2f∆p or ∆p= 2fv_p=A/2πf²). This yields a unidirectional bell-shaped velocity trajectory and monotonic unidirectional displacement (Benson et al., 1989; Grabherr et al., 2008). After each motion, participants indicated direction of perceived motion using buttons held in their hands. Prior to the next stimulus presentation, a minimum 3 second pause was provided to minimize potential motion aftereffects (Crane, 2012a, 2012b). Stimulus size was determined using a symmetric four down/one up (4D/1U) staircase in which peak stimulus amplitude was decreased after four correct responses in a row and increased after each incorrect answer. An initial 2D/1U staircase was implemented in which stimulus amplitude was halved until the first reversal. After this initial staircase, standard PEST rules were implemented to select step size (Taylor & Creelman, 1967).

Six independent thresholds were assessed in order to assess perception predominantly mediated by each peripheral end-organ pair and to assess canal-otolith integration. Specifically, 1 Hz inter-aural y-axis translation (“y-translation”) and 1 Hz superior-inferior z-axis translation (“z-translation”) were assessed with predominant utricle and saccule contributions, respectively (Agrawal et al., 2013; Benson et al., 1986; Kobel, Wagner, & Merfeld, 2021; Kobel et al., 2024). Additionally, 2 Hz yaw rotations and 2 Hz earth-horizontal rotations (i.e., “tilts”) in the approximate plane of the vertical canals (right-anterior left-posterior; RALP, left-anterior right-posterior; LARP) were completed to assess horizontal and vertical canal responses, respectively (Kobel et al., 2023; Wagner, Kobel, & Merfeld, 2022). For RALP tilt and LARP tilt thresholds, six participants (1 middle-aged adult, 5 older adults) were unable to complete all 100 trials of the threshold measures at the displacement limits of the device. Lastly, 0.5 Hz roll tilt (i.e., medio-lateral earth-horizontal rotations about a naso-occipital axis) thresholds were assessed which predominantly reflect integration of vertical canal and otolith cues (Lim et al., 2017; Wagner, Kobel, & Merfeld, 2022).

2.3. General Cognitive and Spatial Navigation Assessments

2.3.1. Self-Administered Gerocognitive Exam (SAGE)

The SAGE is a brief pen and paper self-administered cognitive screening instrument designed to identify MCI and early dementia. The SAGE assesses six broad domains including language, reasoning, visuospatial skills, executive function, memory, and orientation (Scharre et al., 2014). Subjects were provided with Form 1. The participants completed a non-scored portion which included items querying basic demographics. From this section, education level was recorded. The maximum score is 22 and a score between 17–22 is considered within the normal range (Scharre et al., 2010). Most of the subjects completed the SAGE while in the laboratory, however, some completed the SAGE at home and provided the experimenters with the hardcopy for scoring.

For cognitive testing performed in the laboratory, the majority of subject performed this on a separate test day in which they also completed additional balance and gait testing including the Gait Disorientation Test (discussed below). For participants who completed cognitive testing on the same day as vestibular thresholds, this was completed prior to vestibular threshold assessments to minimize potential fatigue.

2.3.2. Digit Symbol Substitution Test (DSST)

The DSST is a widely administered pen and paper measure which assesses multiple cognitive domains including executive function, information processing speed, and working memory (Jaeger, 2018). The subject is provided with a key of numbers and corresponding symbols and a test section with numbers and empty boxes. For the assessment, the participant is asked to fill as many empty boxes as possible with the corresponding symbol that matches the provided number. The score is the number of correctly provided symbols in 90 seconds.

2.3.3. Trail Making Test A and B (TMT-A & TMT-B)

The TMT is a pen and paper test which has been widely used to assess visuospatial ability, processing speed, and mental flexibility. Standard administration was completed in line with previously presented guidelines (Strauss et al., 2006). In the TMT-A, participants are asked to draw a line in order to connect a series of number in consecutive order (1, 2, 3, etc..). The TMT-A reflects attention, visual search, and processing speed. In the TMT-B, participants are asked to connect a series of alternating numbers and letters (1, A, 2, B, 3, C, etc.). The TMT-B examines executive function (i.e., mental flexibility), attention, visual search, and processing speed(Bowie & Harvey, 2006). A cutoff time of 300 seconds was designated to discontinue test administration, however, none of the subjects reached this limit. The time in seconds to complete the task was recorded as the primary outcome variable.

2.3.4. The Gait Disorientation Test (GDT)

The GDT was assessed in 45 participants. Originally proposed by Grove, et al. (2021), the GDT reflects path integration, one of the components of spatial navigation. The GDT was completed on a separate test day, and five participants (2 middle-aged; 3 older) participants did not complete due to scheduling constraints and/or equipment unavailability.

The GDT assesses the difference in time to walk 20 ft (6.096 m) with the eyes open versus eyes closed. These were taken from the functional gait assessment (FGA) item 1 (eyes open) and item 8 (eyes closed), which was administered by a trained, licensed physical therapist (ARW). Participants performed the GDT in their preferred flat soled shoes along a dedicated path in a quiet 30 ft (9.144m) long hallway. The beginning and ending boundary of the 20 ft (6.096 m) pathway was defined by markers fixed to the floor. Before each task, standardized verbal instructions were provided. All participants walked at their preferred pace prior to entering the measurement path and continued walking until instructed to stop after the ending point. Participants were video recorded during testing and times to complete were verified to ensure that only initial and final steps of the trial were included.

2.4. Statistical Analysis

To examine intra-group differences in demographics (i.e., sex, education), chi-squared tests were completed. Consistent with past studies (Bermúdez Rey et al., 2016; Grabherr et al., 2008), vestibular perceptual thresholds displayed a log-normal distribution, thus thresholds were transformed prior to all analyses. All cognitive measures (i.e., DSST, GDT, TMT-A, TMT-B, GDT) were not normally distributed (Shapiro-Wilk <0.001) nor log-normally distributed (Shapiro-Wilk <0.05), thus untransformed values were included in all analyses and non-parametric analyses were used when appropriate. In those adults that were able to complete RALP and LARP tilt thresholds (n=44), both thresholds were highly correlated (r²=0.726, p<0.001) and not significantly different from each other (t=−1.340, p=0.906), thus averages of both thresholds were included to represent vertical canal sensitivity.

Relative relationship between age to each threshold and each cognitive measure was assessed using Spearman’s rank correlation coefficients (Supplemental Digital Content Table 1). P-values were corrected using the Bonferroni multiple comparison adjustment and for all analyses presented p-values in text and tables represent corrected p-values (i.e., p x N). Corrections for multiple comparisons were made across all measures for correlations assessing impact of age (i.e., N=9). A corrected p-value of less than or equal to 0.05 was considered statistically significant.

To explore the relationships between each vestibular threshold and cognitive measures, Spearman’s rank correlation coefficients were assessed (Table 2). For correlation analyses and separate linear regressions, corrections for multiple comparisons were only made across each outcome measure (i.e., for each cognitive assessment; N=5) rather than across the large number of comparisons made (i.e., across all cognitive measures; N=20). While this increases the likelihood of a Type I error, as this was an exploratory study to inform focus of future large-scale studies, we aimed to minimize the likelihood of a Type II error. As well, this study was designed to determine relative strength of vestibular thresholds to assess which modalities (i.e., predominantly canal or otolith) are associated with different cognitive domains.

Table 2.

Spearman rank correlation coefficients for each log-transformed vestibular threshold and assessments of cognition and path integration. Corrected p-values are presented due to multiple comparisons. Corrections were performed across each response variable (i.e., each assessment) using the Bonferroni method to account for multiple comparisons (i.e., p = p *5). Statistical significance (i.e., p≤0.05) indicated with bolded font. RALP/LARP tilt thresholds represent the average of both thresholds for those participants (n=44) that could complete. For GDT, the correlation to RALP/LARP tilt thresholds includes 40 participants who completed both assessments. DSST=Digit Symbol Substitution Test; GDT=Gait Disorientation Test; LARP =left anterior, right posterior; RALP=right anterior, left posterior; TMT-A=Trail Making Test Part-A; TMT-B=Trail Making Test Part-B.

	DSST n=50		TMT-A n=50		TMT-B n=50		GDT n=45
	rs	p	rs	p	rs	p	rs	p
Yaw rotation	−0.228	0.574	0.238	0.499	0.205	0.784	0.298	0.234
RALP/LARP tilt	−0.553	0.001	0.336	0.138	0.374	0.068	0.466	0.012
Roll tilt	−0.537	0.001	0.355	0.062	0.369	0.045	0.161	0.999
Y-translation	−0.359	0.057	0.358	0.058	0.306	0.161	0.416	0.023
Z-translation	−0.754	<0.001	0.547	<0.001	0.606	<0.001	0.278	0.321

Multivariable regression models were used to determine the association between each threshold measure (i.e., yaw, RALP/LARP tilt, roll tilt, y-translation, and z-translation) to each cognitive assessment while controlling for demographic variables including age, sex, and education level. Six adults could not complete RALP/LARP tilt threshold measures, thus, these full regression models only included those with full data sets (N=44) for each cognitive assessment (Table 3). Standardized beta coefficients (i.e., how many standard deviations the dependent variable will change, per standard deviation increase in the predictor variable) are reported for all analyses due to the differences in units/scales for the variables of interest. Residuals of these multivariable regression analyses were normally distributed as assessed by visual inspection of Q-Q plots and Shapiro-Wilk tests (p>0.05) after log-transformation of thresholds. Vestibular thresholds are significantly correlated with each other (r² =0.33–0.62); mean variance inflation factors (VIF) for full models were 2.65 and ranged from 1.92–4.13 indicating moderate correlation.

Table 3.

Multivariable regression models for participants with complete datasets. Six participants could not complete RALP/LARP tilt thresholds, thus are excluded from all analyses. Five participants did not complete GDT testing due to scheduling constraints, one of which also could not complete RALP/LARP thresholds, yielding 40 participants with both GDT and RALP/LARP tilt thresholds. Statistical significance (i.e., p≤0.05) indicated with bolded font. DSST=Digit Symbol Substitution Test; GDT=Gait Disorientation Test; LARP =left anterior, right posterior; RALP=right anterior, left posterior; TMT-A=Trail Making Test Part-A; TMT-B=Trail Making Test Part-B.

	DSST (n=44)			TMT-A (n=44)			TMT-B (n=44)			GDT (n=40)
	βstand	t-stat	p	βstand	t-stat	p	βstand	t-stat	p	βstand	t-stat	p
Age	0.003	−3.25	−0.46	0.064	1.92	0.043	0.002	3.33	0.77	0.049	2.04	0.22
Sex	0.623	−0.5	−1.79	0.05	−2.05	−1.14	0.775	0.29	1.68	0.516	0.66	1.81
Education
College	0.169	1.41	15.61	0.566	0.58	0.97	0.947	0.07	1.2	0.798	0.26	2.19
Graduate	0.135	1.53	16.39	0.721	−0.36	−0.59	0.844	0.2	3.42	0.817	0.23	1.91
Yaw rotation	0.772	−0.29	−1.12	0.197	−1.32	−0.82	0.735	0.34	2.11	0.65	0.46	1.34
RALP/LARP tilt	0.812	0.24	1.25	0.255	1.16	0.98	0.666	−0.44	−3.67	0.555	−0.6	−2.38
Roll tilt	0.668	−0.43	−2.34	0.981	−0.02	−0.027	0.673	−0.43	−3.74	0.528	−0.64	−2.65
Y-translation	0.17	1.4	4.57	0.044	2.1	1.04	0.323	−1	−5.28	0.535	−0.63	−1.56
Z-translation	0.09	−1.75	−5.84	0.064	−1.19	0.043	0.002	1.44	0.77	0.283	1.09	2.80

Separate linear regression models for each threshold and each cognitive assessment were also completed adjusting for age. These individual models allowed assessment of associations for those with missing RALP/LARP tilt thresholds, which represented those adults with the highest thresholds. Residuals for each regression demonstrated normality (Shapiro-Wilk>0.05) after log-transformation of thresholds. As sex was a significant predictor for GDT (Table 3; discussed below), adjustments for sex and age were made for GDT. Demographics (e.g., education level) were not considered in univariate regression analyses due to lack of statistical significance in full regression models. For individual regression analyses, as for correlation analyses, p-values were corrected using Bonferroni adjustment method (i.e., p= p x 5). STATA 17 statistical software was used for all analyses (StataCorp, College Station, TX, USA).

3. Results

Demographic information of included participants is in Table 1. No adults met the criteria of mild cognitive impairment (MCI) on the basis of the SAGE scores (i.e., score≥17). There was not a significant difference between number of females between targeted age groups (Χ²=0.74, p=0.69) nor was there a difference in highest education level reported between the age groups (Χ²=4.61 p=0.330).

3.1. Effect of Age

A significant correlation with age was seen for all measure except yaw rotation thresholds (Supplemental Digital Content Table 1; Supplemental Digital Content Figures 1 and 2) suggesting age-related declines in vestibular perception and measures of cognitive function. A strong correlation with age was seen for both DSST (r_s =−0.774, p<0.001) and z-translation thresholds (r_s=0.731, p<0.001), while all other measures –except yaw rotation--exhibited moderate correlations with age (i.e., r_s =0.54–0.69).

3.2. Associations between Vestibular Thresholds and DSST

Correlation analyses revealed a moderate negative correlation between DSST and RALP/LARP tilt (r_s=−0.53, p=0.001) and roll tilt thresholds (r_s =−0.55, p=0.001) and a strong correlation to z-translation thresholds (r_s=−0.754, p<0.001; Table 3; Figure 1). This suggests that lower DSST scores (i.e., decreased performance) were associated with higher vestibular thresholds (i.e., decreased vestibular sensitivity).

Figure 1.

Scatter plots showing the association between each vestibular threshold and DSST for each participant (n=50). DSST=Digit Symbol Substitution Test; LARP =left anterior, right posterior; RALP=right anterior, left posterior.

No vestibular perceptual thresholds showed a significant association with DSST scores in the multivariable regression model controlling for sex, education level, and all other thresholds. Only a significant association between DSST and age was identified. (Table 3). In individual regression models, after adjusting for age, only z-translation thresholds were negatively associated with DSST scores (β_stand=−0.732, p=0.025; Table 4).

Table 4.

Results of individual linear regression models adjusting for age. Six participants could not complete RALP/LARP thresholds, thus these results reflect data from 44 participants. Five participants did not complete GDT testing due to scheduling constraints, one of which also could not complete RALP/LARP thresholds, yielding 40 participants with both GDT and RALP/LARP tilt thresholds. All p-values are corrected across each response variable (i.e., each cognitive assessment) using the Bonferroni method to account for multiple comparisons (i.e., p=p/N; N= 5). Bolded values indicate significance in age adjusted regression analyses. DSST=Digit Symbol Substitution Test; GDT=Gait Disorientation Test; LARP =left anterior, right posterior; RALP=right anterior, left posterior; TMT-A=Trail Making Test Part-A; TMT-B=Trail Making Test Part-B.

	βstand	t-stat	p-value
DSST (n=50)
Yaw rotation	−0.051	−0.499	1.00
RALP/LARP tilt	−0.099	−0.690	1.00
Roll tilt	−0.258	−2.477	0.085
Y-translation	0.061	0.522	1.00
Z-translation	−0.377	−2.932	0.025
TMT-A (n=50)
Yaw rotation	−0.023	−0.176	1.00
RALP/LARP tilt	−0.167	−0.900	1.00
Roll tilt	−0.022	−0.156	1.00
Y-translation	0.001	0.004	1.00
Z-translation	0.118	0.668	1.00
TMT-B (n=50)
Yaw rotation	−0.026	−0.225	1.00
RALP/LARP tilt	−0.108	−0.716	1.00
Roll tilt	0.081	0.662	1.00
Y-translation	−0.088	−0.671	1.00
Z-translation	0.219	2.073	0.044
GDT (n=45)
Yaw rotation	0.191	1.314	0.980
RALP/LARP tilt	0.299	1.638	0.550
Roll tilt	−0.023	−0.151	1.00
Y-translation	0.469	2.883	0.030
Z-translation	−0.317	−1.748	0.440

3.3. Associations between Vestibular Thresholds and TMT

A moderate positive correlation between z-translation thresholds was identified between TMT-A (r_s=0.547, p<0.001) and TMT-B (r_s=0.606, p<0.001) (Table 2; Figure 2 & 3), suggesting that decreased TMT performance (i.e., longer time to complete) was associated with decreased vestibular sensitivity to superior-inferior translations (i.e., higher thresholds). A statistically significant weak correlation between roll tilt thresholds and TMT-B (r_s=0.369, p=0.045) was also identified.

Figure 2.

Scatter plots showing the association between each vestibular threshold and TMT-A for each participant (n=50). LARP =left anterior, right posterior; RALP=right anterior, left posterior; TMT-A=Trail Making Test Part A.

Figure 3.

Scatter plots showing the association between each vestibular threshold and TMT-B for each participant (n=50). LARP =left anterior, right posterior; RALP=right anterior, left posterior; TMT-B=Trail Making Test Part B.

In multivariable regression models (Table 3), a significant association between age and both TMT-A and TMT-B performance was identified. Z-translation thresholds also showed a significant positive association with TMT-B (β_stand=0.771, p=0.002). No significant association between any vestibular threshold and TMT-A was seen. Similarly, in age-adjusted univariate models (Table 4), vestibular thresholds were not found to have a significant association with TMT-A scores. For TMT-B, only z-translation thresholds were significantly associated with TMT-B (β_stand=0.613, p=0.044).

3.4. Associations between Vestibular Thresholds and GDT

For GDT, a moderate positive correlation between scores and RALP/LARP tilt (r_s=0.466, p=0.012) and y-translation (r_s=0.416, p=0.023) was identified (Table 2; Figure 4). This suggests that decreases in path integration performance (i.e., increased GDT scores) was associated with decreased vertical canal sensitivity.

Figure 4.

Scatter plots showing the association between each vestibular threshold and GDT for each participant (n=45). GDT=Gait Disorientation Test; LARP =left anterior, right posterior; RALP=right anterior, left posterior.

A significant relationship was identified between GDT scores and y-translation thresholds in multivariable regression models (Table 3). In individual regression models adjusting for age and sex only y-translation thresholds were significantly associated with GDT scores (β_stand=0.355, p=0.030; Table 4).

4. Discussion:

Overall, this study further demonstrates associations between age-related changes in vestibular sensitivity and cognition in healthy adults and also adds to evidence that suggests an increased influence of the otoliths on cognition (P. F. Smith, 2019). We emphasize that these findings were true even when measured in a healthy adult population without evidence of vestibular pathology and screened to exclude the potential impact of other health conditions. Past evidence linking cognition to vestibular function has largely been seen in patients presenting with pathologic vestibular loss (Bosmans et al., 2021; Dobbels, Peetermans, et al., 2019; Felfela et al., 2024; Schöne et al., 2024) while only large scale studies have identified significant links in healthy aging populations (Bigelow et al., 2015; Semenov et al., 2016). As well, unlike earlier studies that primarily utilized clinical measures assessing vestibular reflexes, the vestibular measures reported herein - perceptual thresholds - were acquired using similar methodologies for all vestibular thresholds. In this study, overarching correlations between all vestibular thresholds and cognitive assessments were not noted, suggesting modality specific associations between age-related increases in vestibular perceptual thresholds (i.e., decreased vestibular sensitivity) and age-related declines in cognition. This is further bolstered by cognitive associations with perceptual thresholds having predominant otolith contributions (i.e., z-translation, y-translation) noted in age-adjusted regression analyses.

Across general cognitive assessments (DSST, TMT-A & TMT-B), a relationship to z-translation thresholds, a metric predominantly mediated by the saccule, was consistently identified in correlation and regression analyses. The exception being a lack of relationship between TMT-A and z-translation thresholds in age-adjusted regression analyses. Path integration (i.e., GDT) and y-translation thresholds, a metric with predominant utricular contributions, were also significantly associated in correlation and regression analyses.

Additionally, correlation analyses, which by definition did not correct for age, suggest a potential link between vertical canal perceptual sensitivity to general cognition and path integration. RALP/LARP tilt thresholds, reflecting predominantly vertical canal function (Wagner, Kobel, & Merfeld, 2022), were positively correlated with DSST and GDT performance. As well, roll tilt thresholds, reflecting integration of vertical canal and otolith input (Lim et al., 2017; Wagner, Kobel, & Merfeld, 2022), were correlated with DSST and TMT-B. As these associations were not seen in regression analyses, this suggests that these correlations may have captured age-related changes to both. This finding should not be surprising, as age-related factors outside of vestibular degradation are likely to also influence cognitive performance. However, future studies should further explore potential links to determine if a relationship between vertical canal sensitivity and different domains of cognition exist.

As discussed in Methods, corrections for multiple comparisons were made across each outcome variable (i.e., for each cognitive assessment) for both correlation and linear regression analyses. Of note, the majority of the identified statistically significant relationships for correlation analyses (Table 2) remain significant regardless of correction method. Correlations suggesting a weaker relationship (i.e., r_s <0.5) between threshold measures and cognition fail to retain significance with a more stringent correction for multiple comparisons (N=20). As well, this more conservative correction for multiple comparisons would render all individual linear regression analyses to not be statistically significant (Table 4). However, these threshold measures were still found to display the strongest relative relationship with cognitive function, with standardized beta coefficients being found to be ~1.4–2.7x larger than the next highest weighted threshold. As this study was designed to determine relative strength of vestibular thresholds to various cognitive domains, this suggests that regardless of correction, we would still be able to identify these relationships between cognition and specific measures of vestibular function.

Our results suggesting a link between age-related changes in otolith function and general cognition are largely consistent with past population-based studies of aging (i.e., BLSA, NHANES). These earlier studies identified age-related changes in balance performance and cVEMPs which were associated with decreased performance on the DSST and TMT-B (Semenov et al., 2016). Moreover, mediation analyses in both of these studies revealed that assessments reflecting vestibular function (i.e., balance performance or cVEMP) mediated ~5–11% of the association between aging and cognitive performance (Bigelow et al., 2015; Semenov et al., 2016).

In this current asymptomatic population, we observed an association between path integration, as measured by the GDT, and thresholds reflecting perception with predominant utricular (i.e., y-translation) and vertical canal (i.e., RALP/LARP tilt) contributions in age-adjusted analyses, which is consistent with the well documented vestibular contributions to path integration (Adamo et al., 2012; Anson et al., 2019; Bent et al., 2004; Grove et al., 2024; Xie et al., 2017). The GDT has been found to be sensitive to acute (i.e., 4–8 week post-surgically induced loss) and chronic (i.e., 3 month to 20 year) peripheral vestibular dysfunction (Grove et al., 2021). While no participants in this dataset exhibited scores indicative of potential peripheral vestibular hypofunction (i.e., GDT ≥ 4.5s), we were still able to identify changes in path integration performance, as quantified by the GDT, which correlated to vestibular perceptual sensitivity.

Past studies have identified a more consistent link between path integration in older adults to measures of the saccule (i.e., cVEMPs) or while relationships to measures of utricular integrity (i.e., oVEMP) and canal function (i.e., vHIT) have been less consistently identified. For example, Anson et al. (2019) found a significant association between cVEMPs and path integration performance, as assessed by the Triangle Completion Test (TCT), while associations between TCT and oVEMPS and horizontal canal vHIT were not identified using data from the BLSA. However, Xie et al. (2017) identified in older adults, both with and without vestibular dysfunction, a significant relationship between TCT to both cVEMP and horizontal canal vHIT. Conversely, Grove et al., (2024), using a larger sample from the BLSA, only identified a relationship between TCT performance and oVEMPs, and no relationship between cVEMP and vHIT were seen.

We did identify a link between our measure of utricular sensitivity (i.e., y-translation thresholds), in line with results from the TCT in Grove et al., 2024. However, did not identify a link between our measure of saccular sensitivity (i.e., z-translation thresholds) to GDT performance, in contrast to Anson et al. (2019) and Xie et al. (2017) which may reflect differences in samples and test procedures. These past studies linking vestibular function to path integration quantified vestibular reflexive function abnormalities categorically (i.e., normal, unilaterally absent responses, bilaterally absent responses) (Anson et al., 2019; Grove et al., 2024; Xie et al., 2017) while our perceptual measures were continuous. As well, the TCT requires participants to complete the final leg of a triangle and return to start after being passively guided on the first two legs (Adamo et al., 2012; Anson et al., 2019; Xie et al., 2017). Thus, the TCT represents a more complex spatial navigation task and previously, path integration task complexity has been found to influence links to vestibular function (Péruch et al., 2005). As well, the TCT requires the participant to make turns, providing canal excitation, while the GDT only requires walking in a straight line. As such, different peripheral vestibular input is provided during path integration tasks, potentially influencing the associations between thresholds and task performance that we identified. Of note, the GDT requires walking 6.096 m which may allow participants to achieve steady state, while the TCT requires participants to only walk a short distance (i.e., 2.12 m), thus, is largely capturing gait initiation and cessation. Past research implementing galvanic vestibular stimulation (GVS) have identified phase dependent contributions of vestibular function to gait, which are modulated by gait stability (Dakin et al., 2013; Forbes et al., 2017; Magnani et al., 2021). Our findings suggesting utricular contributions to path integration as quantified by the GDT in contrast to predominantly saccular contributions to the TCT may reflect the unique contributions of the vestibular system to different phases of gait which are emphasized by the different walking lengths.

Functional vestibular loss secondary to aging has been well-documented (e.g., Agrawal et al. 2009, 2012, 2019; Peterka et al. 1990a, 1990b; Serrador et al. 2009). A larger degree of age-related changes in measures reflecting otolith function in comparison to canal function, as assessed by both reflexive and perceptual measures, have also been identified which may influence the ability to assess relationships between modality specific changes in vestibular function and cognition. High-frequency VOR function reflecting both the horizontal and vertical canals, as assessed by vHIT, is relatively stable until 70–80 years of age (Agrawal et al., 2012; Li et al., 2015; McGarvie et al., 2015) and significant age-related decreases in peak caloric responses have not been consistently identified (Mallinson & Longridge, 2004; Peterka et al., 1990b). However, age-related changes in otolith function as assessed by VEMPs, begin in after approximately 50–60 years of age (Agrawal et al., 2012; Chang et al., 2012; Iwasaki et al., 2008; Layman et al., 2015; Rosengren et al., 2011; Singh & Barman, 2016). As well, across modalities (i.e., translation, rotation), perceptual thresholds have been reported to be stable until approximately 40 years of age (Bermúdez Rey et al., 2016). However, vestibular thresholds with predominant inputs from the otoliths have consistently shown age-related effects (Agrawal et al., 2013; Bermúdez Rey et al., 2016; Bremova et al., 2016; Karmali et al., 2017; Kingma, 2005, p. 20; Roditi & Crane, 2012); an effect of aging on thresholds with predominant horizontal semicircular canal contributions (i.e., yaw rotation) has not been as consistently identified (Karmali et al., 2017; Kobel, Wagner, Merfeld, et al., 2021; Roditi & Crane, 2012; Seemungal, 2016).

This pattern of preferential age-related changes on measures of otolith function was also identified in this study as a significant correlation to age was identified for all measures except yaw rotation thresholds while the largest age-related effects were seen for z-translation thresholds. In our data, a consistent link between z-translation thresholds and cognitive measures was also identified, even in analyses correcting for age, suggesting that these associations were not solely capturing the larger impact of age on z-translation thresholds. As well, a similar magnitude of age-related change in RALP/LARP tilt thresholds was identified relative to z-translation thresholds; however, significant associations between perception with predominant contributions from the vertical canals and cognitive performance was not identified, suggesting unique links between perceptual measures with predominant saccular contributions and cognitive performance. However, we acknowledge that a potential common cause impacting both z-translation thresholds and cognition cannot be ruled out. Specifically, cross sectional links between cVEMPs and hippocampal volume could represent a common mechanism (i.e., hippocampal atrophy) influencing both cognition and saccular function. As well, while there has been a consistent link noted between measures of otolith function to cognitive performance across multiple clinical and pre-clinical populations (P. F. Smith, 2019), the preferential impact of age must be considered when assessing age-related changes in vestibular function and cognition.

Links between vestibular function, cognition, and balance have been proposed in part due to the increased incidence of vestibular dysfunction (Harun et al., 2016; Wei et al., 2018) and falls (Allali et al., 2017; Allan et al., 2009) in adults with impaired cognition. Recent evidence suggests that in patients with MCI and AD, saccular integrity, as quantified by cVEMP amplitude, and VOR function, reflecting integrity of the horizontal canal, were associated with postural sway and likelihood of falls over a one year period (Biju et al., 2022). As past research suggests a consistent link between vestibular perceptual thresholds and quiet stance balance performance in cognitively healthy adults (Bermúdez Rey et al., 2016; Karmali et al., 2021; Wagner et al., 2021; Wagner, Kobel, Tajino, et al., 2022), future studies should further examine the relationships between vestibular perception, cognition, and balance.

There are several limitations to this study exploring potential relationships between vestibular function and cognition. The cross-sectional nature of this study does not allow casual inferences between age-related changes in vestibular function and cognitive performance. Future longitudinal studies and/or intervention studies assessing both vestibular perception and cognition should be completed to assess causality. Previous work by Sugaya et al. demonstrated improvement in TMT performance, and therefore presumably visuospatial cognition, attention, and executive function, in 60 patients with chronic dizziness after completing vestibular rehabilitation (Sugaya et al., 2018). Similarly, we have previously shown that perceptual learning can induce short-term improvements in vestibular perception (Wagner, Kobel, Tajino, et al., 2022); future studies should aim to assess if these improvements are correlated to changes in cognition both short-term and long-term, which would also contribute to an assessment of causation.

Additionally, hearing loss has also demonstrated consistent associations to changes in cognition in multiple populations (Dobbels, Peetermans, et al., 2019; Lin, 2011; Lin et al., 2013; P. F. Smith, 2021) and due to their shared anatomical location, vestibular loss and hearing loss often co-occur (Lucieer et al., 2016; Zingler et al., 2007). A large body of preclinical animal work focused on disambiguating auditory and vestibular contributions to cognitive decline suggests links between vestibular function and cognition independent of auditory dysfunction (P. F. Smith, 2021). However, human studies rigorously assessing both vestibular loss and hearing loss and links to cognitive performance are limited, but suggest potential interactions between vestibular and auditory dysfunction (Bosmans et al., 2020, 2022; Dobbels, Mertens, et al., 2019; Gommeren et al., 2023). The present study did not adjust for hearing loss; thus, future studies should explore this further. As well, while several past studies have demonstrated that perceptual thresholds are sensitive to peripheral end-organ damage (Agrawal et al., 2013; Bremova et al., 2016; Kobel et al., 2023; Priesol et al., 2014; Valko et al., 2012), our study did not include traditional reflexive testing of peripheral end-organ function. Future studies should aim to include traditional assays in order to allow correlation between reflexive and perceptual measures and to more definitively identify end-organ specific contributions.

Finally, our population included is not representative of the larger demographics of the aging population, in part due to our intentionally stringent inclusion criteria to reduce the impact of other chronic health conditions and focus on healthy aging. Our identification of correlations between cognition and vestibular function are found in this population – absent any diagnosed dysfunction – are perhaps the most important contribution of this manuscript.

Only 3 out of 50 participants did not have a college degree or higher and 22 participants (44%) had completed an advanced graduate degree. However, despite our inclusion of well-educated and ultra-healthy aging adults, mean DSST and TMT scores were similar to past large-scale population based studies (Bigelow et al., 2015; Semenov et al., 2016). While future studies should assess if similar relationships hold in a larger scale population, this strengthens evidence that these findings are related largely to age-related changes in function.

Supplementary Material

Supplemental Figure 1

Supplemental Digital Content Figure 1. Scatter plots depicting the association between each cognitive assessment and age. SDC_Fig1.pdf

Supplemental Figure 2

Supplemental Digital Content Figure 2. Scatter plots depicting the association between each vestibular threshold assessment and age. SDC_Fig2.pdf

Supplemental Table 1

Supplemental Digital Content Table 1. Correlation coefficients between age and each vestibular and cognitive measure. SDC_Table1.pdf