The Effect of Hearing Loss on Cognitive Function in Subjective Cognitive Decline Open Access

The aging process affects both cognition and hearing. Increasing evidence has demonstrated that age-related hearing loss (HL) is a major risk factor for incident dementia [1-5]. According to the recent report, elimination of all modifiable risk factors for dementia might prevent or delay up to 40% of worldwide dementias [6]. HL is particularly important in terms of its prevalence and high relative risk for dementia [6, 7]. The authors documented evidence for hearing-related cognitive decline in mice and suggested that early HL can be one of the determinants between physiological and pathological cognitive aging [8, 9].

With an aging population and abundant information on dementia, an increasing number of people report a feeling of cognitive alterations and actively seek medical aid. Out of them, individuals without objective cognitive impairment are categorized under a specific clinical entity, subjective cognitive decline (SCD). Heterogeneous etiologic factors contribute to the feeling of cognitive decline: the aging process, depression, anxiety, personality, physical health concerns, the side effects of medication, and sleep disturbances [10, 11]. The criterion for SCD is a self-experienced persistent cognitive decline with normal performance on standardized cognitive tests for mild cognitive impairment (MCI) or Alzheimer’s disease (AD) [10, 12, 13]. Because the boundary between SCD and MCI is a gray zone, a clear-cut threshold for objective cognitive impairment for MCI has not been defined. Practically, deficits of >1.5 standard deviation (SD) on a single test are considered to exclude a diagnosis of SCD [13]. A higher risk of progressive cognitive deterioration has been demonstrated in cognitively unimpaired persons with SCD compared with cognitively unimpaired persons without SCD [11]. SCD can lead to the clinical manifestation of dementia over a 10-year period in some individuals. That is, SCD may represent the late phase of preclinical AD [10, 12, 14]. Individuals perceive a decline in cognition as early as 5 years before the onset of MCI [15]. The speed of cognitive deterioration and severity of clinical symptoms depend on the individual’s brain reserve. Individuals with more cognitive reserve can process tasks more efficiently with the same number of synapses. Or they may have an enhanced ability to recruit additional cognitive networks in response to increasing task difficulty [16, 17]. Because cognitive decline in patients is roughly equivalent to increased task difficulty in normal individuals, more cognitive reserve could buffer functional deterioration in the presence of brain pathology.

Despite the prevalence and clinical importance of both HL and SCD, few reports are available on the relationship between hearing and SCD. Recent studies showed that “self-reported” HL is related to higher risk of incident SCD [18, 19]. Further, we questioned whether “audiometric” HL influences cognitive performances in our SCD cohort. In this study, we planned to compare various cognition-related indexes between the normal-hearing and hearing-impaired SCD individuals.

We used the baseline data of a multicenter cohort study named the CoSCo (a co hort study to identify predictors for the clinical progression to MCI or dementia from s ubjective co gnitive decline) study that monitors SCD longitudinally to determine the predictive factors for cognitive deterioration. Participants were enrolled from six centers between November 2018 and November 2019. For this study, those who met the inclusion criteria and underwent the hearing test were grouped by the hearing status. The institutional review board of each center approved the study and the study conformed to recognized standards.

The Seoul Neuropsychological Screening Battery 2nd version (SNSB-II), a form of norm-referenced assessment, was employed. It is composed of comprehensive cognitive tests that evaluate five different cognitive domains: attention, language, visuospatial, memory, and frontal/executive functions [20]. Nine subtests used in this study included the Digit Span Test, Korean-Boston Naming Test, Rey Complex Figure Test: Copy, Rey Complex Figure Test: Delayed recall, Seoul Verbal Learning Test-Elderly’s version: Delayed recall, Digit Symbol Coding, Controlled Oral Word Association Test, Korean-Trail Making Test-Elderly’s version: Part B time, and Korean-Stroop Color Word Test. Each test score was converted to the z-score ([test score − mean]/SD) based on the age-, sex-, and education-stratified norms (the mean and SD of a normative sample) presented in the SNSB-II.

The inclusion criteria were: (1) subjective complaints of cognitive decline, (2) ≥60 years old, (3) education period of ≥6 years, (4) deficits of 1.5 SD or less (≥7 percentile) on all cognitive tests, and (5) the score of a memory test (Seoul Verbal Learning Test-Elderly’s version: Delayed recall) in the range of −1.5 to 0 SD (7–50th percentile). We named it amnestic SCD, which is more likely to progress to MCI/AD earlier and thus shortens the overall study period. Individuals showing a deficit of >1.5 SD (<7th percentile) below the normal mean on any cognitive test were excluded because of the possibility of MCI. These schemes are illustrated in Figure 1. Those with brain or medical diseases that may affect cognition, uncontrolled psychiatric conditions, and alcohol/drug dependence were excluded. Demographic data (age, sex, education, and occupation), medical history (hypertension, diabetes, and depression), and lifestyle factors (alcohol, smoking, and obesity) were collected at baseline. For apolipoprotein E (APOE) genotyping, genomic DNA was extracted from the venous blood, and the genotype was determined by one-stage polymerase chain reaction [21]. Global cognition was evaluated using the Korean-Mini-Mental State Examination (K-MMSE) [22]. The participants who met the inclusion criteria underwent pure tone audiometry. A hearing assessment was performed in a soundproof chamber using a calibrated audiometer. Air-conduction thresholds were measured from 0.25 to 8 kHz with headphones in a well-standardized manner. A pure tone average (PTA) of thresholds at 0.5, 1, 2, and 4 kHz was computed. The participants were grouped into the normal-hearing (NH) (PTAs of ≤25 dB in both ears) and HL (PTAs of >25 dB in both ears) groups. Those with unilateral HL were excluded from the analysis because various ways of auditory processing are possible according to the laterality and degree of HL.

The participants answered questionnaires. The validated Korean version of the hearing handicap inventory for the elderly (K-HHIE) was used to verify subjective HL [23, 24]. The activity of daily living was evaluated by the Korean-Everyday Cognition (K-ECog) scale composed of 39 items about memory, language, visuospatial, and executive functions [25].

For quantitative assessment of cerebral amyloid beta (Aβ), florbetaben (18F) positron emission tomography (PET) was taken and processed by the precedent method [26]. The PET data were co-registered to the individual three-dimensional T1-weighted MR images. The regional standardized uptake values (SUVs) for the MRI-based, predefined regions of interest were obtained. The regional SUV ratio (SUVR) was calculated by dividing the SUV of the region of interest by that of the cerebellar cortex. The global SUVR was yielded by averaging 90 regional SUVRs.

Oligomeric Aβ in the plasma was measured according to the protocol described in the earlier literature [27]. Multimer Detection System-Oligomeric Aβ was quantified using the inBlood^TM OAβ test (People Bio Inc., Gyeonggi-do, Korea).

Multicentered brain MRI was processed for volumetric measurement using a deep learning-based whole brain segmentation tool. AQUA 2.0 software (Neurophet Inc., Seoul, Korea) was employed for the MRI processing pipeline and normative data analysis. The detailed methods appeared in the previous publication [28, 29]. Obtained subregional gray matter volumes were adjusted according to the intracranial volume normalization approach as suggested elsewhere [30]. Then, the corresponding percentile of an individual was calculated by quantile regression on the adjusted subregional volume with covariates of age and sex based on the East-Asian dataset described previously [28]. Finally, the normative percentiles of subregional volumes were compared between the groups using the T test.

Group differences were examined for the demographic factors, APOE4 genotype, medical history, lifestyle factors, K-MMSE, K-HHIE, K-ECog, quantified Aβ biomarkers, and cognitive z-scores. SPSS 18.0 (SPSS Inc., Chicago, IL, USA) was used for statistical analyses at a significance level of 0.05. P values were determined by the Mann-Whitney test and χ² test for continuous and categorical variables, respectively. Because age was the only demographic factor that was statistically different between the groups, probably age-dependent variables such as K-MMSE, K-HHIE, and K-ECog were adjusted for age by the analysis of covariance.

A total of 120 participants were recruited. This study included 102 participants: 57 in the NH and 45 in the HL groups. Eighteen with unilateral HL were excluded from the analysis. Mean PTAs of the right and left ears in the NH group were 16 and 17 dB, respectively; in the HL group, 40 and 44 dB, respectively (Fig. 2). The demographic and clinical characteristics at baseline are listed by the hearing status in Table 1. Between the groups, only the age was different statistically. The mean age of the NH/HL group was 68.3/74.0 years (p < 0.001).

The MMSE scores representing global cognition were similar between the groups. The K-HHIE for subjective HL was understandably higher in the HL group (p < 0.001). There was no group difference in the K-ECog total score (Table 2). The Aβ biomarker indexes (amyloid PET SUVR and plasma oligomeric Aβ) showed no statistical differences between the groups (Table 3).

Out of nine neuropsychological tests, the z-score of the Stroop Color Word Test was significantly lower in the HL group than in the NH group: mean (SD) of −0.06 (0.87)/0.27 (0.73) in the HL/NH group (p = 0.025) (Table 4). In brain volumetry, the HL group exhibited reduced cortical gray matter volumes in four subregions: left temporal pole, left caudal middle frontal gyrus, left hippocampus, and right isthmus of the cingulate gyrus compared with the NH group (Fig. 3).

This study investigated whether audiometric HL affected cognition-related indexes in a multicenter SCD cohort. For the cognitive performance of neuropsychological tests to be compared between the NH and HL groups, the age and education level should be controlled first [6, 17, 31]. Therefore, preadjusted z-scores were obtained with reference to age-, sex-, and education-stratified norms of the neuropsychological test battery used in this study. And because no group differences were found in the indexes related to dementia risk factors including APOE4, education, occupation (social activities), hypertension, diabetes, depression, alcohol, smoking, and obesity [6, 7], potential confounders were thought to have been controlled. Global cognition based on the MMSE was similar between the groups within the normal range. The activity of daily living/functioning assessed with the Everyday Cognition scale was also similar between the groups. The AD biomarkers measured by both the amyloid PET SUVR and plasma Aβ oligomerization did not differ between the groups.

Since the definition of SCD requires normal cognitive testing, none of the participants showed objective cognitive impairment (Fig. 1). However, comprehensive neuropsychological assessments revealed that hearing-impaired SCD participants performed worse on the Stroop Color Word Test [32] with a significantly lower mean z-score than the normal-hearing counterparts. For the standard Stroop Color Word Test, participants were asked to name the ink color in which a conflicting color name was presented (e.g., the word red in blue ink) [33, 34]. It is not clear why hearing-impaired participants performed worse only in the Stroop test but not in the other tests. A possible explanation is that the robust Stroop effect, one of the most reliable phenomena in cognitive science, could discriminate subtle cognitive alterations in SCD participants who had normal objective cognitive function [35, 36]. The Stroop task is a particularly useful measure of the executive aspects of attentional control. Selective attention is needed to suppress the more automatic processing of word reading and enhance the less automatic processing of color naming [37]. The word-color interference is expressed as a response time delay for resolution of the conflict by attentional selection in naming the incongruent color [33, 35, 38]. The areas of brain activation during this task depend on the experimental setting. The anterior cingulate and prefrontal cortices have been most reported to participate in the Stroop performance [33, 34, 36, 38, 39]. The anterior cingulate cortex has been regarded as a central executor that coordinates and integrates the activity of multiple attentional subsystems [40].

The potential mechanism for the link between peripheral HL and cognitive decline is that when listening is effortful, more resource-demanding auditory processing taxes cognitive networks, and consequently, less cognitive reserves lead to accelerated cognitive decline [7, 41-43]. The term compensation means the use of atypical networks in the face of standard network pathology with resultant functional reorganization [16, 17]. Applying these hypotheses to the Stroop task performance in the hearing-impaired SCD subjects, it is supposed that compensatory auditory processing caused by difficult perception in the auditory networks can deplete a limited pool of resources allocated to executive attentional networks. As a result, fewer resources would be available for the greater task demand on the Stroop test [44, 45]. An fMRI study provided evidence for functional reorganization of brain networks in long-term sensorineural HL. Intra- and inter-network functional connectivity has altered across the sensory and higher order cognitive networks [46]. Another plausible mechanism is reduced cognitive stimulation due to HL. The brain has the capacity to respond to environmental enrichment or stimulation mainly by dendritic branching, which results in increased cognitive reserve [47]. Hearing-impaired SCD individuals may carry out cognitive and social activities less frequently than normal-hearing counterparts. High-level engagement in social activity and large social networks are related to better cognitive function in later life [48]. A community-based cohort study has demonstrated that early-life education, midlife occupational attainment, and late-life mentally stimulating leisure activities are cognitive reserve-enhancing factors which reduce the risk of dementia [31]. In short, HL may exert a direct adverse effect on cognitive performance through compensatory auditory processing consuming cognitive resources or an indirect negative effect on cognitive function through reduced cognitive reserve-enhancing factors. However, given the nature of a cross-sectional study, the possibility of reverse causality could not be completely ruled out.

Our brain volumetric analysis exhibited four areas of significantly reduced gray matter volumes in the HL group: the left temporal pole, left caudal middle frontal gyrus, left hippocampus, and right isthmus of the cingulate gyrus. Notably, the middle frontal gyrus is a part of the prefrontal cortex, and the hippocampus is a major component involved in memory. However, a cause-and-effect relationship between the hearing status and brain volumetry is controversial and needs further investigation. HL and brain volume reduction may be more likely the products of a common etiology such as accelerated aging in susceptible individuals [45].

This is the first investigation that focused on the hearing-related cognitive performance in SCD. It was built on detailed demographic, physical, biological, and neuropsychological data collected from multiple large centers. Hearing levels were audiometrically documented, and multiple cognitive domains were assessed. However, this study has limitations. First, since we have no healthy controls without SCD and HL, the Stroop task performance could not be compared among the control, NH SCD, and hearing-impaired SCD. Second, it is a baseline cross-sectional study. We plan to follow up the cognitive performance further in both hearing groups of SCD.

HL exerted an adverse effect on cognitive function associated with the Stroop Color Word Test. HL might primarily affect prefrontal executive function, specifically selective attention required for the demanding Stroop task. HL was also related to gray matter volume reductions in a few brain subregions including the temporal pole, caudal middle frontal gyrus, hippocampus, and isthmus of the cingulate gyrus, although further investigation is needed for the causal relationship. This study provides new evidence for a potential link between hearing and cognition in SCD, an emerging clinical entity. Because the SCD population is probably at a higher risk for future dementia, management of HL would be more important in this population.

The authors thank Seong A. Shin, PhD, who has made the figure of brain volumetry.

This study was conducted ethically in accordance with the World Medical Association Declaration of Helsinki. The study protocol was reviewed and approved by the Institutional Review Board of The Catholic University of Korea St. Mary’s Hospital, approval number, KIRB-20180807-026. Written informed consent was obtained from participants to participate in the study.

The authors have no conflicts of interest to declare.

The project has been funded by a grant from the Ministry of Health and Welfare (HI18C0530), Republic of Korea.

Dong Won Yang and Shi Nae Park: conception and design, data acquisition and interpretation, drafting and revising, and final approval of the manuscript. Seong Hee Ho, Yun Jeong Hong, Jee Hyang Jeong, Kee Hyung Park, SangYun Kim, Min Jeong Wang, and Seong Hye Choi: data acquisition. So Young Park: data analysis and interpretation and writing the manuscript. Regina E.Y. Kim: MRI processing and brain volumetric analysis. Dong Won Yang: obtained funding.

The data collected and generated during the current analysis are available from the corresponding author for researchers who meet the criteria for access on reasonable request.