Abstract
The vestibular system is an important contributor to balance control, spatial orientation, and falls risk. Recent evidence has shown that Alzheimer’s disease (AD) patients have a higher prevalence of vestibular impairment relative to healthy controls. We sought to evaluate whether vestibular loss is specifically associated with poor spatial cognitive skills among patients with mild cognitive impairment (MCI) and AD. We enrolled 50 patients (22 MCI and 28 AD) from an interdisciplinary Memory Clinic and measured vestibular physiologic function in all patients. Spatial cognitive function was assessed using the Money Road Map Test (MRMT) and the Trail Making Test Part B (TMT-B). General cognitive function was assessed with the Mini-Mental Status Examination (MMSE). In multivariable linear regression analyses adjusted for age, gender, education level, and MMSE, MCI and AD patients with vestibular loss made significantly more errors on the MRMT relative to patients with normal vestibular function (β= 7.3, 95% CI 2.4, 12.1 for unilateral vestibular loss and β= 6.4, 95% CI 1.9, 10.9 for bilateral vestibular loss). We further stratified AD patients into “spatially normal” and “spatially impaired” groups based on MRMT performance, and found that the prevalence of vestibular loss was significantly higher in the spatially impaired AD group relative to the spatially normal AD group. These findings support the hypothesis that vestibular loss contributes specifically to a decline in spatial cognitive ability in MCI and AD patients, independently of general cognitive decline, and may predict a “spatially impaired” subtype of AD.
INTRODUCTION
Patients with Alzheimer’s disease (AD) have an increased risk of imbalance, falls, and spatial disorientation (e.g., wandering behaviors) [1–3]. These symptoms are thought to contribute to the increased morbidity [4, 5], earlier institutionalization [6, 7], and shorter survival [8] observed in patients with AD relative to cognitively-normal adults. The vestibular (inner ear balance) system is an important contributor to balance control, spatial orientation [9–11], and falls risk [12–15], and recent evidence suggests that AD patients have greater levels of vestibular impairment relative to age-matched controls [16].
There is increasing recognition that AD patients have heterogeneous clinical features [17–19]. Some patients with AD may experience greater motoric and spatial impairments early in the disease course, such as impairments in gait, spatial memory, and spatial navigation [20–22]. The reasons for this clinical heterogeneity and the relatively poorer motor and spatial function in some AD patients is unknown. The vestibular system provides critical inputs to the brain about the orientation of the head with respect to gravity, and vestibular function is integral to spatial cognitive processes such as spatial memory and spatial navigation. At present, it is unknown whether vestibular impairment might be associated with greater spatial impairment among AD patients. In this study, we sought to evaluate whether AD patients with poorer spatial cognition were more likely to have vestibular impairment relative to AD patients with better spatial cognitive function.
In this study, we evaluated 50 patients with both mild cognitive impairment (MCI) and AD who were seen in the Johns Hopkins Memory and Alzheimer’s Treatment Center (JHMATC), an interdisciplinary memory disorder clinic, and the Johns Hopkins Alzheimer’s Disease Research Center (JHADRC). We used the Money Road Map Test (MRMT) as our primary measure of spatial cognitive function, and also administered the Trail Making Test Part B (TMT-B) which taps into spatial cognition. We evaluated the association between vestibular function and MRMT and TMT-B performance within the MCI and AD patients. We also classified AD patients into “spatially normal” and “abnormal” subgroups based on their MRMT score, and compared the prevalence of vestibular impairment across these two groups. Our findings provide evidence that vestibular loss may specifically contribute to the onset of a “spatial” subtype of AD. As a corollary, screening for and treating vestibular loss (e.g., with vestibular rehabilitation) in MCI and AD patients may forestall, mitigate or prevent the evolution of these symptoms.
METHODS
Study participants
This was a single-center prospective cross-sectional study. Participants were recruited from the JHMATC and the JHADRC from December 2014 until February 2017. Inclusion criteria for the study were 1) Age≥55 years; 2) Diagnosis of MCI or AD; 3) MMSE score≥11; 4) Fluency in English; 5) Ability to obtain informed consent from the participant or a legally authorized representative. A diagnosis of MCI or AD was made using the National Institute on Aging-Alzheimer’s Association diagnostic criteria [23, 24]. Prior to participation, patients with an MMSE score≥11 were assessed by a Memory Center physician to determine their ability to followexamination procedures.
Patients were excluded from the study if they had a previous history of vestibular disease, were unable to understand examination procedures, or if they could not participate in study procedures due to blindness, poor neck range of motion, or cervical instability. Demographic characteristics (age, gender, and education), and MMSE score were obtained from the patients’ charts. Education was classified as less than high school, high school, college, or greater than college. All participants provided written informed consent. The study was approved by the JohnsHopkins Institutional Review Board.
Vestibular testing
Vestibular function was assessed with the sound-evoked cervical vestibular evoked myogenic potential (cVEMP), a measure of saccular function. Recordings were made with a commercial electromyographic system (software version 21.1; Natus Neurology, Middletown, WI), with self-adhesive electrodes from GN Otometrics (Schaumburg, IL). Testing methods have been published in detail previously and are described briefly here [25–27]. Participants sat on a chair with their heads inclined at 30° from the horizontal. Trained examiners placed recording electromyographic (EMG) electrodes on the sternocleidomastoid (SCM) muscle and the sternoclavicular junction bilaterally. A ground electrode was placed on the manubrium sterni. Patients were instructed to turn their heads against resistance to activate the SCM muscle. A sweep of auditory tone bursts (500 Hz, 125 dB SPL) was delivered monaurally through headphones (VIASYS Healthcare, Madison, WI), and inhibitory potentials were recorded from the ipsilateral SCM muscle. The presence or absence of a cVEMP response was recorded for each ear, indicating normal or impaired saccular function, respectively, as characterized by published guidelines [25–27].
Visuospatial testing
Money Road Map Test (MRMT)
The MRMT assesses visuospatial ability, and has been used extensively in patients with mild cognitive impairment and dementia [28]. The procedure has been published in detail, and will be briefly explained here [28–30]. In this test, the participant was shown a 2D representation of a small city map on a sheet of paper. A walking path was drawn on the city map with 32 turns. The subject was asked to imagine that they were travelling along the route through the city. As the examiner traced the route, the examiner asked the subject at each turn whether a right or left turn would be required to continue along the route. The map was oriented in a fixed position in front of the patient during the tracing of the route and the subject was not permitted to turn the sheet of paper. The participant’s responses were recorded at each turn by the examiner up until a 300 second time limit. The main outcome of interest was the number of errors (i.e., incorrect responses) out of 32 turns on MRMT. Four patients (1 MCI and 3 AD) did not complete the test as their time exceeded the predetermined 300 second time limit. In order to compute a score for these participants, any turns that were not completed within the 300 second time limit were considered incorrect in the analyses.
Trail Making Test-B (TMT-B)
The TMT-B assesses executive function, set-shifting, attention, processing speed, and visual scanning ability. The procedure has been published in detail, and will be briefly discussed here [31, 32]. In this test, the subject was shown 26 encircled numbers and letters scattered across a piece of paper. The subject was asked to connect a series of letters and numbers in alternating consecutive order (1, A, 2, B, 3, C, etc.) using a pen, given the location of the encircled “1” as the starting point. The outcome of interest was the number of errors (i.e., incorrect lines drawn) out of a potential 25 lines. One subject did not complete the TMT-B within 300 seconds. Any lines that the subject did not attempt within the 300 second time limit were considered incorrect in the analyses.
To address a potential bias in our method of determining errors on MRMT and TMT-B imposed by the time limits, a second method of determining errors on MRMT and TMT-B was utilized. In this method, the number of errors a subject made was divided by the number of turns attempted within 300 seconds to calculate the percent correct. The percent correct was multiplied by 32 for MRMT or 25 TMT-B to extrapolate the number of errors the subject would have made if they had been allowed to finish the test without a time limit. In sensitivity analyses described below, we used this second method of calculating error rate in regression analyses.
Statistical analysis
Descriptive analyses were conducted to examine participant demographics, MMSE score, and errors on visuospatial tests by category of cognitive impairment (MCI and AD).
We next evaluated the relationship between visuospatial ability and vestibular function in multivariable analyses. Visuospatial ability was estimated using the MRMT and TMT-B error rates, and vestibular function was categorized as normal function, unilaterally absent function, or bilaterally absent function. Multivariable linear regression models adjusting for demographic characteristics (age, gender, and education) were developed to explore the association between vestibular loss and visuospatial impairment in all patients. To evaluate whether the association between vestibular function and visuospatial ability differed based on diagnosis, we performed separate regression analyses in MCI and AD patients. Further, we added MMSE to the multivariable regression models to assess whether any relationship between vestibular function and visuospatial ability may be explained by a general decline in cognitive function rather than a specific association with visuospatial ability. Wald F-tests were performed to assess if there were significant differences between the regression coefficients of unilateral vestibular loss and bilateral vestibular loss.
Finally, to explore the hypothesis that vestibular loss contributes to a “spatial” subtype of AD, we dichotomized AD patients into a “spatially normal” and a “spatially impaired” group based on their MRMT error score. Previously published studies in healthy adults have demonstrated that normal adults make a mean 2 or less errors on the MRMT [30, 34], so≤2 errors on MRMT was used as a cutoff for the spatially normal phenotype. We performed Fisher’s exact test to compare the prevalence of vestibular impairment in each group. All analyses were performed using STATA version 14 (College Station, TX, USA).
RESULTS
Characteristics of sample
Fifty patients (22 MCI and 28 AD) completed vestibular testing and one or both of the visuospatial tests (Table 1). Among these, 49 patients (22 MCI and 27 AD) were able to complete the MRMT and 40 patients (20 MCI and 20 AD) were able to complete the TMT-B. There was a higher percentage of females in the AD group (75%) compared the MCI group (45%). MCI patients were slightly younger with a mean age of 73.1 (SD = 8.8) years, compared to AD patients, who had a mean age of 76.3 (SD = 6.7) years. A majority of MCI patients (77.3%) and AD patients (64.3%) attained an education level of college or greater. MMSE was greater in MCI patients (Mean = 26.5, SD = 2.2) than in AD patients (Mean = 21.5, SD = 4.2). On the MRMT, MCI patients made fewer errors (Mean = 4.8, SD = 6.2) compared to AD patients (Mean = 10.1, SD = 7.0). Similarly, MCI patients made fewer errors on the TMT-B (Mean = 1.6, SD = 5.0) compared to AD patients (Mean = 2.8, SD = 3.2).
Demographic characteristics of study population and performance on visuospatial tests
a27 AD completed MRMT. b20 MCI completed TMT-B. c20 AD completed TMT-B.
MRMT regression analysis
Multivariable linear regression analyses were conducted to assess the association between vestibular loss and errors on MRMT in MCI and AD patients after adjusting for potential confounders (Table 2). Relative to patients with normal vestibular function, more errors on MRMT were made by patients with unilateral vestibular loss (β 7.7, 95% CI 2.9, 12.4) and with bilateral vestibular loss (β 6.9, 95% CI 2.5, 11.2). Females made 6.4 more errors on the MRMT than males (95% CI 2.1, 10.6). Compared to participants with less than a high school education, higher levels of education were significantly associated with fewer errors on the MRMT (High school β –11.9, 95% CI –21.8, –2.1; College β –11.9, 95% CI –21.4, –2.4; Greater than college β –12.7, 95% CI –21.9, –3.5).
Vestibular loss and number of errors on Money Road Map Test (MRMT)
aNormal vestibular function used as reference level. bLess than high school used as reference level.
To assess whether vestibular loss-associated decline in visuospatial ability were different between MCI and AD patients, analyses were conducted on MCI and AD groups separately and interaction terms were considered. In MCI patients alone, unilateral vestibular loss, but not bilateral vestibular loss, was significantly associated with a higher number of errors on the MRMT (β 6.4, 95% CI 0.02, 13.7). In AD patients alone, unilateral vestibular loss (β 11.2, 95% CI 2.8, 19.6) and bilateral vestibular loss (β 9.6, 95% CI 1.8, 17.4) were significantly associated with a higher number of errors on the MRMT. However, interaction terms for vestibular loss category (unilateral and bilateral) by diagnosis (MCI/AD) were not significant (data not shown), indicating no statistical difference in the association between vestibular and visuospatial function by diagnosis. According to Wald F-test, there was no significant difference in mean number of errors on the MRMT between unilateral and bilateral vestibular loss in regression analyses of all three groups (i.e., combined, MCI and AD; data not shown).
TMT-B regression analysis
Multivariable linear regression analyses were performed to assess the association between vestibular loss and errors on TMT-B in MCI and AD patients after adjusting for potential confounders (Table 3). Compared to patients with normal vestibular function, individuals with bilateral vestibular loss made significantly more errors on the TMT-B (β 3.7, 95% CI 0.3, 7.1). Patients with unilateral vestibular loss made a non-significant higher number of errors on the TMT-B (β 2.0, 95% CI –1.7, 5.7). The association between vestibular loss and errors on the TMT-B did not persist when MCI and AD groups were examined separately.
Vestibular loss and number of errors on Trail Making Test-B (TMT-B)
aNormal vestibular function used as reference level. bLess than high school used as reference level.
To address the potential bias that the 300 second time limit imposed on participants for the MRMT and TMT-B, the alternative method of calculating errors was used, whereby the percent errors made within 300 seconds was used to extrapolate the number of errors the subject would have made if no limit were imposed. The results of the regression analyses using the alternative method of calculating errors were not substantially different from the results presented in Tables 2 and 3.
Adjusting for general cognitive ability
To evaluate the hypothesis that vestibular loss is associated with visuospatial impairment independent of a general decline in cognition, MMSE score was added to the regression models for MRMT and TMT-B as a proxy for general cognitive ability in MCI and AD patients (Table 4). The addition of MMSE in the model for MRMT did not substantially impact the coefficients for vestibular loss or of other variables in the model. Specifically, after adjusting for MMSE, MCI and AD patients with unilateral vestibular loss (β= 7.3, 95% CI 2.4, 12.1) and bilateral vestibular loss (β= 6.4, 95% CI 1.9, 10.9) made significantly more errors on the MRMT relative to patients with normal vestibular function. MMSE was not significantly associated with errors on MRMT (β –0.2, 95% CI –0.7, 0.3).
Vestibular loss, MMSE, and errors on the Money Road Map Test (MRMT) and Trail Making Test-B (TMT-B)
aNormal vestibular function used as reference level. bLess than high school used as reference level.
In contrast, when MMSE score was added to the regression model for TMT-B, there was no longer a significant association between bilateral vestibular loss and TMT-B errors (β 2.3, 95% CI –1.3, 6.0, p = 0.194). In this model MMSE had a borderline non-significant association with errors on TMT-B (β –0.5, 95% CI –1.0, 0.05, p = 0.075).
Prevalence of vestibular loss in “spatial” versus “non-spatial” subtypes of AD
To explore the hypothesis that AD patients with impaired visuospatial ability have a greater likelihood of vestibular loss, AD participants in the study were categorized into “spatially normal” and “spatially impaired” based on errors on the MRMT (Table 5). In the AD subgroup with normal spatial cognitive ability (n = 4), 75% of patients (n = 3) had normal vestibular function, 25% of patients (n = 1) had unilateral vestibular loss, and 0% (n = 0) of patients had bilateral vestibular loss. Interestingly, a 25% prevalence of vestibular impairment was also observed in a prior study of healthy older adults [16]. In the AD subgroup with impaired spatial cognitive ability (n = 23), 4.4% (n = 1) had normal vestibular function, 26.1% (n = 6) had unilateral vestibular loss, and 69.6% (n = 16) had bilateral vestibular loss. According to Fisher’s exact test, the prevalence of vestibular loss was significantly higher in the spatially impaired AD group relative to the spatially normal AD group (p = 0.002).
Spatially impaired AD is associated with increased prevalence of vestibular loss
DISCUSSION
In this study of patients with MCI and AD, poorer spatial cognition was significantly associated with reduced vestibular function. Moreover, when AD patients were categorized into “spatially impaired” versus “spatially normal” patients, the subgroup of AD patients with spatial cognitive impairment were significantly more likely to have vestibular loss. Indeed, we observed that AD patients with normal spatial cognition had similar prevalence rates of vestibular loss as previously seen in healthy older adults [16]. These findings support the hypothesis that vestibular loss contributes specifically to a decline in spatial cognitive ability in patients with MCI and AD. Our study builds on an emerging body of evidence that the vestibular system is critical to spatial cognitive function [35, 36]. Animal experiments have shown that peripheral vestibular ablation in rats results in impaired spatial memory andnavigation skills as assessed with a foraging task [37]. Human studies have also shown that bilateral vestibular deafferentation to treat bilateral vestibular nerve tumors results in impaired spatial navigation skills as well as reduced hippocampal volumes relative to age-matched controls [11]. The hippocampus is thought to play a critical role in spatial memory and navigation via transmission through hippocampal place cells that encode a cognitive map [38]. The peripheral vestibular organs makes substantial projections to the hippocampus [39–43]. Moreover, a further study in rats demonstrated that vestibular signaling is critical for the normal functioning of hippocampal place cells [44]. These studies offer compelling evidence for an association between vestibular loss and spatial cognitive decline. Indeed, some have hypothesized that vestibular loss contributes to the onset of AD more broadly [45], given the large population of cholinergic fibers that project from the vestibular system to the hippocampus. However, at this time, a link between vestibular loss and a spatial subtype of AD may have a stronger evidence base.
Several studies suggest the presence of a subtype of AD characterized by disproportionately greater impairment in spatial cognitive function[17, 47]. These patients tend to have a more rapidly progressive disease course, and have characteristic PET imaging findings of reduced metabolism in the posterior parietal cortex (responsible in part for spatial cognition) relative to AD patients without a chief spatial deficit [21, 48]. The phenotypic heterogeneity of AD may reflect different pathophysiologic mechanisms, with vestibular loss potentially contributing to a spatial variant of AD.
In our study, we specifically examined the association between saccular function and spatial cognition, since we previously found that of the vestibular impairments reduced saccular function had the strongest association with visuospatial ability [9, 49]. The saccule is the vestibular end-organ involved in detecting the orientation of the head with respect to gravity, and is thought to play a pre-eminent role in the orientation and encoding of space. Stimulation of the saccule has been shown to result in activation of vestibular cortical areas as measured with EEG [50, 51]. Saccular stimulation has also been shown to activate multisensory vestibular cortex involved in spatial processing in several fMRI studies [52, 53]. In prior work, we observed associations between saccular function and spatial cognitive skills in healthy older adults [9, 49]. Additionally, we found that patients with AD specifically had poorer saccular and utricular function relative to age-matched controls [16]. The saccule appears to have particular relevance for spatial cognitive function, and may thus represent a therapeutic target for improving spatial cognition.
In this study, we administered both the MRMT and TMT-B as measures of spatial cognition in the MCI and AD patients [30, 54–57]. The MRMT primarily assesses mental rotation and spatial navigation, while the TMT-B assesses a range of cognitive domains including spatial ability, executive function, set-shifting, attention, visual scanning ability, and motor skills [56, 57]. We observed a significant association between vestibular loss and poorer MRMT performance that persisted even after adjustment for general cognitive status (i.e., MMSE). This suggests that vestibular loss was associated with spatial cognition independent of overall cognitive level, and also that the MRMT and the MMSE tap into discrete cognitive skills. In contrast, we observed a weaker association between vestibular loss and poorer TMT-B performance that was no longer significant with the addition of MMSE to models. This may reflect the broader cognitive skills measured by the TMT-B which overlapped with the domains measured by the MMSE.
We note important limitations of this study. This was a cross-sectional study. As such the results are correlational and we cannot draw a causal inference about the role of vestibular loss in spatial impairment in MCI and AD from these analyses alone. We acknowledge the potential for reverse causality, or alternatively that other factors such as traumatic brain injury, hypertension, diabetes, and/or cardiovascular disease may contribute to both vestibular loss and spatial cognitive decline [45]. It is also possible that brain regions involved in spatial cognition co-localize with central saccular afferent processing, such that atrophy of brain regions involved in spatial cognition may also cause saccular dysfunction. It has also been suggested that there may be bidirectional interactions between cognitive state and vestibular sensitivity [55]. Nevertheless, our findings can be interpreted in the context of previous experiments that suggest a causal contribution of vestibular loss to spatial cognitive dysfunction. Another limitation of our study was the 300 second time limit imposed on the MRMT and TMT-B, which was established for practical reasons to limit total study time for participants. Notably, only four participants did not complete the MRMT and only one participant did not complete the TMT-B within the established time limit. We utilized two different methods to extrapolate the number of errors made by these patients, with similar results. Additionally, the paper and pencil tests of spatial cognition employed in this study may have limited real-world applicability. Further work should assess the ecological validity of these tests, and evaluate whether vestibular loss may mediate functional declines in MCI and AD patients associated with poor spatial cognition. Furthermore, given recent work suggesting altered tuning of cVEMP responses in older adults [58], and advantages of utilizing multi-frequency VEMPs [59], future studies should address another limitation of the current study by adding tone burst stimulation at 750 and 1000 Hz, in addition to 500 Hz. Lastly, we did not differentiate between subtypes of MCI, which may or may not be of likely AD cause. Further work focused on spatial cognition in MCI patients could restrict the study sample to MCI patients with likely AD etiology.
In summary, we found that vestibular loss is significantly associated with poorer spatial cognition in MCI and AD patients, and may predict a “spatially impaired” subtype of AD. Future investigations should evaluate whether vestibular loss is associated with behaviors consistent with impaired spatial cognition such as getting lost, wandering, and falls, which are associated with higher morbidity and mortality [4, 8] and earlier institutionalization [6, 7]. Additionally, future studies may be warranted to evaluate whether vestibular therapy improves spatial cognitive skills and clinical outcomes in patients with MCIand AD.
Authors’ disclosures available online (https://www.j-alz.com/manuscript-disclosures/17-0751r2).
Footnotes
ACKNOWLEDGMENTS
EO was funded by a NIA/NIH K23 Award (1K23AG043504–01), the Roberts Gift Fund, and the Ossoff Family Fund. AH was funded by a NIH T32 Award (5T32DC000027–25) and an AAO-HNSF Core Grant (349386). YA was funded by a NIH K23 Award (5K23DC013056–02). The Alzheimer’s disease Research Council is supported by an NIA award (3P50AG005146–32S1).