Dementia poses a global burden—a 2022 report estimated that in 2019, 55 million people were living with a diagnosis of dementia (Gauthier et al., 2022). Future projections suggest that dementia will continue to increase in prevalence, approaching 140 million individuals by 2050 (Gauthier et al., 2022). To slow this exponential growth, novel interventions, either improving management of those already diagnosed, or via the prevention of those at an increased risk of developing it, are imperative.

Logically, risk factors for impaired cognition precede those for dementia, due to its clinical progression. Notably, age is distinguished as an important risk factor—the natural aging process leads to an inherent risk of cognitive decline, independent of dementia. Indeed, aging itself is simultaneously associated with a 70% increased incidence of dementia (Juan and Adlard, 2019). In addition, it is well established that loneliness and social isolation are contributing factors in poor cognitive performance, and are associated with an increased rate of decline (Cacioppo and Hawkley, 2009; Dominguez et al., 2021). Further to this, research has shown that individuals categorized as socially isolated have a 50% increased relative risk of developing dementia (Evans et al., 2018; Dominguez et al., 2021). Interestingly, a recent Lancet review (Livingston et al., 2020) focusing on various modifiable risk factors for the development of dementia, showed the importance of risk reduction, suggesting that modifying such risk factors has the potential to reduce the likelihood of dementia by 40% (LaPlume et al., 2022). Among different risks factors, including traumatic brain injury, hypertension, depression, and diabetes mellitus, the review highlights hearing loss (HL) as one of the potential factors that, when excluded, reduced the risk of dementia by 8%. Mild, moderate or severe HL particularly in the middle-life (specified as between the ages of 45 and 65) has been associated with an increase of 2, 3 and 5 times respectively in the risk of reduced cognition and dementia (Lin et al., 2011; Livingston et al., 2020). Furthermore, observational studies suggest that the severity of hearing impairment (HI) is associated with a risk of accelerated cognitive decline (LaPlume et al., 2022), and one report even demonstrated that the dementia risk increased linearly with the severity of baseline HL (1.27 per 10 dB loss) (Ford et al., 2018). However, limitations remain regarding confounding factors, which may influence the generalizability of these findings (Ford et al., 2018).

In this review, we explore HL as a risk factor for the development of cognitive impairment and dementia, focusing on molecular mechanisms. Through this, we summarize the research surrounding clinical manifestations of HL in dementia, proposed mechanisms to this relationship, and the role of interventions, including the use of hearing aid (HA) devices to minimize the effects of HL in the progression of cognitive impairment to dementia. We highlight the importance of further understanding the relationships between the proposed mechanisms and cognitive impairment, with the eventual aim of providing effective novel interventions to mitigate the risk of HL in the development of dementia.

Hearing loss affects currently 466 million people Worldwide (World Health Organisation, 2021). Individuals suffering from disabling HL account for over 5% of the global population (World Health Organisation, 2021) and it is predicted that by 2050, nearly 1 in every 10 people will require hearing rehabilitation. Causative factors of hearing loss include genetic factors, ear infections, cerumen impaction (impacted ear wax), trauma to the ear or head, loud noise/loud sounds (NHL), ototoxic medicines and others.

HL mechanisms implicated in dementia predominantly surround sensorineural hearing changes, in which there is dysfunction of the cochlea. Most often, it is seen as age-related hearing loss (ARHL) or presbycusis, which affects around 40% of individuals over the age of 65 (Gates and Mills, 2005). Causes of HL including presbycusis, noise-induced hearing loss (NIHL) and ototoxicity, all precipitate permanent HL, and subsequently, result in limited management options (Lee and Bance, 2019).

ARHL is highly polygenic, with over 100 genes known to underlie human non-syndromic hearing impairment (Lewis et al., 2018; Van Camp and Smith, 2023), and of these many genes, possibly each makes small contributions to create an estimated heritability of 36–70% (Nagtegaal, 2019). Most cases of genetic deafness imply alterations of the cochlea, the auditory sensory organ; for instance, the OTOF gene encodes the protein otoferlin, which allows synaptic vesicles to fuse to the plasma membrane in the ribbon synapse. Thus, mutations in this gene can lead to a deficiency in exocytosis in the inner hair cells, which interrupts auditory signal transmission and can cause prelingual deafness (Vona et al., 2020). However, some forms of genetic hearing loss can involve failures in the central auditory system. Genome wide association studies (GWAS) that have focused on the genes associated with ARHL, although showing some overlap, do not show strong agreement with one another (Nagtegaal, 2019; Wells et al., 2019; Liu et al., 2021; Lewis et al., 2022). When considering a genetic relationship between hearing loss and AD, no GWAS have found a direct causal link between ARHL and AD. However, a recent study using UK biobank data (Brenowitz et al., 2020) demonstrated that a genetic risk for AD also influences speech-in-noise hearing. In addition, in another report (Mitchell et al., 2020), individuals with higher polygenic risk score (PRS) for AD were more likely to experience hearing difficulty than those with lower PRS.

Sudden sensorineural hearing loss (SSHL) has also been identified as a risk factor for the development of dementia. This form of HL is characterized as an otological emergency, defined by HL of at least 30 dB over 72 h, affecting at least three different auditory frequencies, often with a viral, vascular, or autoimmune etiology (Lee and Bance, 2019). In a retrospective cohort study comparing the incidence of dementia in individuals with and without a previous diagnosis of SSHL, it was established that the occurrence of SSHL was associated with a 1.39 times higher likelihood in the incidence of all-cause dementia (Tai et al., 2021).

However, specifically within presbycusis, a variety of other mechanisms have also been implicated, including metabolic factors (for instance, mitochondrial dysfunction), oxidative stress (including changes in reactive oxygen species and Superoxide dismutase deficiency), neurotransmitter imbalance (GABA deficiency), among others (Jafari et al., 2021) (Figure 1). Further insight into the underlying mechanisms and how these different classes of ARHL may herald dementia, is not yet completely understood (Bowl and Dawson, 2019).

Poorer scores on the Mini Mental State Examination (MMSE) have been shown to strongly correlate with deficits in audiological testing (Quaranta et al., 2015; Golub et al., 2019; Saji, 2021; Mohammed et al., 2022; Huang et al., 2023). This link is supported by the demonstration of a relationship between HL and MRI brain atrophy (Jafari et al., 2021). Despite a recent study suggesting otherwise (Marinelli et al., 2022), meta-analyses pooling observational studies have further strengthened evidence of an association between ARHL and cognitive decline (Loughrey et al., 2017; Mamo et al., 2018; Liang et al., 2021). Similar correlations between age-associated hearing loss and cognitive decline have been reported in mice (Dong et al., 2018). However, this observational data should be interpreted with caution, as there is a subsequent inability to draw causal links. Additionally, potential confounders that are often seen in the elderly, such as nutritional issues, visual impairment, vascular risk factors, frailty, bad physical health, depression, and other mental illnesses, could contribute to the trend seen (Hirose et al., 2014; Gill et al., 2020). Among those, the weakening of the vascular system is an interesting link, since it is involved in both deafness and neurodegeneration. Vascular networks of the cochlea (Kurata et al., 2016; Nyberg et al., 2019) and the auditory cortex might be impaired in genetic forms of hearing loss, although neural activity by itself can also affect the remodeling of the vascular system (Lacoste et al., 2014; Whiteus et al., 2014).

Interestingly, central auditory pathway disorder in the studied cohorts correlated more with cognitive changes than non-central HL. Why this occurs is not completely understood, but it is hypothesized that the dementia-related neurodegeneration and healthy aging could affect auditory areas of the brain, and lead to subsequent HI (Bidelman et al., 2014; Johnson et al., 2021). Age-related hearing problems are common among people with dementia and are associated with poor cognitive function and reduced quality of life (Maharani et al., 2018a), suggesting that sensory markers could be useful to detect and target cognitive aging.

Contrastingly, there could be an indirect mechanism, as HI predisposes individuals to social isolation and subsequent dementia risk (Sardone et al., 2020). Both dementia and HL present highly heterogeneous conditions, and therefore the investigation of various subtypes within each could be valuable to help further the field of research exploring their interrelationship. Different forms of both dementia and hearing loss may have different trajectories and studies to elucidate this would allow more effective treatments, as shown in frailty (Segaux et al., 2019).

Mirroring human studies, investigations with animal models of AD, including 5XFAD, APP/PS1 and 3XTg transgenic mice have demonstrated both significant HL and degeneration of spiral ganglion cells in the cochlea (Wang and Wu, 2015, 2021; O'Leary et al., 2017; Liu et al., 2020; Weible et al., 2020). Additionally, 5XFAD mice subjected to surgical deafness through tympanic membrane resection i.e., conductive hearing loss (CHL), showed greater cognitive impairment than those with no HI (Kim et al., 2020). This could imply a reciprocal relationship—that AD can lead to auditory deficits and auditory deficits can influence cognition. Whether these AD models can be applied to dementia generally, is not yet known.

Animal models attempting to elucidate the relationship between dementia and HL center around rodents with induced hearing deficits. Two main paradigms are currently utilized to induce these deficits. Most common methods applied expose rodents to high-volume background noise for several days, inducing permanent trauma (Wang and Wu, 2015; Zhuang et al., 2020; Kurioka et al., 2021a; Li et al., 2021; Paciello et al., 2021), or by the bilateral cochlear ablation. Alternatively, HI is mechanically modeled through insertion of a silicone mold into the ear canal, followed by permanent suturing to close the canal (Paciello et al., 2021). Additionally, administration of ototoxic substances such as furosemide and kanamycin can be employed to induce sensorineural HL, but this is not used often (Shen et al., 2021). These methods inflict physical trauma and stress and subsequently, could influence cognition via mechanisms independent of HI, confounding results. Following HL induction, mice undergo acoustic brainstem response testing, a quantitative method of measuring inner ear and hearing pathway responses to sound (Zhuang et al., 2020; Li et al., 2021). Animal studies more specifically looking at brain region histology conduct further posthumous histology of the animal brain (Li et al., 2021).

One of the key mechanisms explored through animal models and human cases is how, following HL, subsequent decline in hippocampal neurogenesis can lead to cognitive deterioration (Kurioka et al., 2021b). The hippocampus, primarily involved in memory and auditory information processing, is one of the key areas affected in dementia. Interestingly, studies in a rat model with hearing loss and/or Aβ administration demonstrated that the group with lower cognitive abilities was the one with both amyloid-beta (Aβ) and SSHL (Chang et al., 2019). In this report, it was demonstrated a significant decrease in hippocampal synaptic proteins in the Aβ-deaf group, implying that HL influences synaptic plasticity, and that there must be a connection between the central auditory cortex and the hippocampus (Chang et al., 2019).

Another potential link between hearing loss and cognitive impairment in animals implies changes in neuroinflammatory markers. Two animal model studies examining NIHL demonstrated alterations in microglia and the presence of irregularly shaped somas when compared to the control mice (Zhuang et al., 2020; Li et al., 2021). Similar results were reported in a conductive hearing loss (CHL) animal model, showing morphological microglial changes specifically in the dentate gyrus and subgranular zone (Kurioka et al., 2021b). Therefore, it appears that microglial activation does occur following HL and could cause a consequent impairment of hippocampal neurogenesis. However, recent studies seem to contradict these findings, showing that there is no correlation between spatial learning ability and the level of hearing loss or altered microglial density in the hippocampus following noise exposure in rats, suggesting that other mechanisms are involved in the hippocampal-dependent cognitive dysfunction due to noise exposure (Patel et al., 2022). Besides changes in glial cell activation, other studies implementing hearing loss by the use of toxins in mouse models of neurodegeneration have shown increases in inflammatory cytokines such as IL-1β and TNF-α (Ren et al., 2011; Shen et al., 2021).

There are also evidences from human and preclinical studies that AD pathological hallmarks, including hyperphosphorylated Tau (p-Tau) and Aβ, could be affected by HL. A recent report in C57Bl6 in mice exposed to noise induced hearing loss have shown reduced cognition and p-tau and lipofuscin in the hippocampus (Park et al., 2018). Another study in rats also demonstrated that chronic noise exposure (CNE) was associated with tau hyperphosphorylation in the hippocampus and the prefrontal cortex (Cui et al., 2012). In agreement with this, human studies have associated ARHL with higher CSF levels (and accelerated rates of elevation) of p-Tau, but have demonstrated no association between ARHL and the levels of Aβ₄₂ in CSF (Xu et al., 2019; Sarant et al., 2022). In line with this, a cross-sectional cohort study utilizing positron emission tomography scans to investigate the presence of Aβ, also found no association between HL and Aβ load (Sarant et al., 2022). In contrast, studies in rats have shown that chronic exposure to noise resulted in increased generation of endogenous Aβ levels in hippocampus (Cui et al., 2015).

In line with reports showing alterations in hippocampal function, HL has been involved in changes on neurotransmitter expression as well. ARHL mouse models have shown changes in NMDA receptor expression within the hippocampus, which can take place after 4 months of natural HL (Cui et al., 2009). Additionally, GABAa and GABAb receptors were altered during this time period. The change in these neurotransmitter receptors is implicated in hippocampal synaptic plasticity, affecting hippocampal learning, and resulting in cognitive decline (Guo et al., 2021).

An emerging hypothesis links social memory impairment and HL—partially due to limitations of social interaction faced by those with a HI (Beckmann et al., 2020). A study using a mouse model of congenital deafness with OTOF gene knockout, attempted to investigate this through eliminating hearing input from the cochlea to the brain, resulting in decreased social memory. If this hypothesis is correct, it would help explain why associated brain regions' functionality, for example, the hippocampus, is affected following HL (Glick and Sharma, 2020). Furthermore, there is the possibility of a compensatory mechanism, whereby the brain may offset certain neural networks to promote better hearing. In line with this, a human MRI study of HI individuals demonstrated that auditory brain regions displayed an increase in activity, whilst other areas showed a consequential decrease (Rigters et al., 2017). Therefore, this could indicate that this activity decrease is linked to the reduction in cognitive function seen in HL.

As discussed, observational studies have established an association between HI and an increased risk of accelerated cognitive decline. The emphasis placed on HL as a modifiable risk factor for dementia (LaPlume et al., 2022) has led to the reasonable hypothesis that audibility restoration could potentially alleviate cognitive decline, by counteracting the postulated mechanisms which underlie this process and by reducing social isolation, loneliness, and depressed mood (Jiang et al., 2023). It is hypothesized that hearing restoration, may reverse cognitive changes of cortical reallocation of sensory processing patterns, leading to consequent cognitive gains (Glick and Sharma, 2020).

In two short-term rodent studies modeling CHL through use of ear plugs, significant cochlear degeneration (Kurioka et al., 2021b) and changes in the neurotransmitter expression within auditory neurons (Kurioka et al., 2020) was demonstrated following reduced auditory inputs. Interestingly, removal of the ear plugs led to recovery of almost all these anatomical changes, suggesting that similar treatment with either HAs or cochlear implantation in humans could reverse such changes, and perhaps cognitive processing.

Overall, many observational, prospective and retrospective cohort, cross-sectional and longitudinal studies have linked HA use to lower rates of cognitive decline (Maharani et al., 2018b; Glick and Sharma, 2020; Cuoco et al., 2021; Fernandes and Mastroianni Kirsztajn, 2021; Sugiura et al., 2021, 2022; Bucholc et al., 2022; Dillard et al., 2022; Vella Azzopardi et al., 2023) and dementia (Maharani et al., 2018b; Mahmoudi et al., 2019; Byun et al., 2022; Dillard et al., 2022; Naylor et al., 2022). Recent data from the UK biobank support these findings, showing no increased risk of dementia in people with hearing loss using hearing aids (Jiang et al., 2023). Strikingly, one of the studies demonstrated that the diagnosis of dementia is associated with a 54% subsequent reduction in HA usage compared to those without dementia (Naylor et al., 2022). The importance is thus placed on not only increasing HA usage amongst older individuals at higher risk of both HL and dementia, but also working to maximize adherence in this community.

On the other hand, two prospective longitudinal studies found no statistical significance between cognitive scores pre- and post- HA fittings (Sarant et al., 2020; Kawade et al., 2022). Furthermore, the only double-blind, randomized controlled trial evaluating the cognitive benefit of HAs in patients with AD also found no statistically significant relationship (Nguyen et al., 2017). There is a clear lack of consensus of the current evidence, but this could be explained by the various limitations and differences across all the study designs.

Generally, studies investigating cognition pre- vs. post- cochlear implantation reported either no change (Sonnet et al., 2017; Kramer et al., 2018; Sarant et al., 2019), an improvement (Jayakody et al., 2017; Claes et al., 2018; Mosnier et al., 2018; Völter et al., 2021, 2022a,b; Calvino et al., 2022), or a mixed picture across the various subtests (Sorrentino et al., 2020; Huber et al., 2021). Cochlear implantation is one phase of a complex rehabilitation process involving multiple appointments and training programmes following surgery (British Cochlear Implant Group, 2023). These all pose opportunities for increased engagement in cognitive stimulation, known to reduce risk of cognitive decline. Whether the cognitive improvement seen across these studies is due to restored hearing itself or a confounding influence of rehabilitation, would be difficult to determine (Völter et al., 2022b).

Much of the heterogeneity across both the hearing aids and cochlear implantation literature exists due to the variety of tests used to measure cognitive function. Some papers choose to use standardized screening tests like the MMSE (Sonnet et al., 2017; Sorrentino et al., 2020; Herzog et al., 2022; Vella Azzopardi et al., 2023), or the MoCA (Cuoco et al., 2021; Fernandes and Mastroianni Kirsztajn, 2021) as an assessment of global cognitive function, whilst other studies focus on assessing subdomains of cognition through use of investigations like the ALAcog (Sarant et al., 2019; Völter et al., 2021; Calvino et al., 2022), ADAScog (Nguyen et al., 2017), Cogstate (Sarant et al., 2019, 2020), RBANS-H (Claes et al., 2018; Brewster et al., 2020; Calvino et al., 2022) and various other subtests (Jayakody et al., 2017; Mosnier et al., 2018; Füllgrabe, 2020; Kurioka et al., 2020). Using tests like the MMSE, which utilize auditory assessments, leaves HI individuals at a direct disadvantage. Cognition prior to rehabilitation could be underestimated, and treatment may simply allow the participant to perform better due to increased ability to understand the task presented (Füllgrabe, 2020). A ceiling effect is also observed through use of MMSE (Sonnet et al., 2017)—those with a normal cognition pre-intervention are unable to show any further improvement. The introduction of a standardized neurocognitive battery, which assesses all necessary subdomains and is adapted for the HI, needs to be implemented to allow comparison across the literature.

Further sources of heterogeneity across results could be due to the varying durations of studies. Undoubtedly, measures of cognitive function are susceptible to time due to the cognitive decline observed with aging. Some studies examined cognitive outcomes over longer periods, ranging from 3.5 to 18 years (Maharani et al., 2018b; Naylor et al., 2022; Sugiura et al., 2022), whilst others observed significantly shorter periods; for example, one report (Glick and Sharma, 2020) studied participants over 6 months, whilst another (Fernandes and Mastroianni Kirsztajn, 2021) over only 12 weeks. Shorter trials should be interpreted with additional caution—interventions may not have significant time to influence a change, nor do they consider the normal cognitive decline represented over time.

Overall, a clear trend between HL and dementia is demonstrated across the literature. However, evidence is largely composed of observational studies and there remains a lack of interventional studies relating HL to dementia. This is due to a clinical dilemma- that it would be unethical to deny treatment to those with HL, especially when considering the ramifications of untreated HL, not just in relation to dementia. The introduction of open label trials, recruiting from surgical candidates for cochlear implantation, could circumvent this. These trials could build on previous limitations by including a standardized neurocognitive battery. Furthermore, despite a multitude of animal models demonstrating potential underlying mechanisms linking HI and cognitive decline, further research, particularly with longer follow-up periods and invasive recording technology is required to help interrogate the mechanisms of action.

Uncertainties also remain in the implementation of HAs and CIs as interventions against cognitive decline. As a result, algorithms and guidelines may need to be synthesized from future interventional studies to best improve patient outcomes. There should be a focus on maximizing early HL diagnoses and swift implementation of auditory treatment to mitigate the risks of cognitive decline. Models calculating dementia risk would allow the stratification of the geriatric population and perhaps help clinicians decide risk and subsequent necessary auditory treatment. Additionally, technology such as smart phone connected HAs could target the current limitations encompassing HA usage in those with dementia.

AA wrote the current understanding of the association between dementia and hearing loss. AJ wrote the introduction and the most of the section of causes of hearing loss and dementia and made the figure. BK wrote the clinical interventions and their limitations. IS wrote the mechanisms underlying hearing loss, cognitive function, and AD Pathology in animal models. MB-H and MS designed, edited the manuscript and organized the references. All authors contributed to the article and approved the submitted version.