The consequences of unmanaged age-related hearing loss are far-reaching (1). In recent years, the connection between hearing loss and cognition has been increasingly investigated (2–4). Hearing loss in midlife has been identified as the most significant modifiable risk factor (alongside high LDL cholesterol) to dementia in later life (5). Accordingly, identifying strategies to prevent or delay the onset of dementia is a central priority within healthy ageing initiatives. Despite this interest, the pathways underlying the hearing–cognition relationship remain poorly understood. Two main hypotheses have been proposed. The first suggests a shared neuropathological origin, such as neurodegeneration or chronic inflammation (6). The second, the “cascade” hypothesis, posits that hearing loss initiates downstream changes that accelerate cognitive decline (7). Social isolation is frequently cited as one such factor because it is independently associated with both hearing loss (8, 9) and cognitive decline (10). However, empirical evidence for this pathway remains limited and inconsistent.

A systematic review of longitudinal studies found that only one out of fifteen formally tested social isolation as a mediator, and it reported no evidence to support this role (11, 12). One reason for this gap may be the inconsistent and often simplistic measurement of social isolation across studies. Some have relied on marital status or frequency of contact as proxies (12), while others have used single-item classifications (13). These approaches fail to capture the multidimensional nature of social isolation, which encompasses structural, functional, and qualitative aspects of social relationships (14, 15). More comprehensive tools, such as the Medical Outcomes Survey (MOS) Social Support Survey, assess multiple domains, such as emotional, tangible, and informational support, and may provide a more valid measure (16).

Prior studies have also differed in their methodological rigor. For example, one study (17) reported that loneliness and social isolation mediated the association between self-reported hearing impairment and episodic memory over 10 years, while another (13) found only partial support using a simplified measure of social isolation. Both studies relied on self-reported hearing loss and narrow cognitive assessments, which may limit causal inference and generalisability. These limitations highlight the need for research using objective hearing measures, validated social isolation tools, and broader cognitive assessments.

The Hertfordshire Ageing Study (HAS) addresses several of these gaps. It includes objective audiometric measures of hearing, a validated multidimensional measure of social isolation, and a general cognitive assessment, all within a longitudinal design where hearing was assessed prior to social isolation and cognition. While the current analyses do not formally test mediation, they examine two key associations that provide foundational evidence for future mediation models:

By clarifying these long-term associations using robust measures, this study contributes to understanding whether social isolation may be a relevant factor in the hearing–cognition pathway.

At timepoint 2, which was conducted in 2003–2005, there was high attrition, resulting in 294 participants who had completed hearing measures at both timepoints. There were 121 participants who had died between timepoint 1 and timepoint 2, and 409 participants who did not respond, declined to participate in timepoint 2, or were untraced. The average age of the study participants was 67 years at timepoint 1 and 76 years at timepoint 2. The majority (58.8%) were men and were current smokers (10.2%) or ex-smokers (53.1%) (Table 1). Most of the participants were married (74.1%), drank 10 units or less of alcohol per week (49%), or no alcohol (34%). Of the clinical diagnoses, high blood pressure was the most prevalent (30.2%), with depression (13.3%) and heart attack (9.6%) the next most prevalent.

When comparing non-responders (those who died before timepoint 2 and those who declined/untraced) to the study participants, the non-responders were generally: older age, male gender, worse average hearing threshold, current smoking status at timepoint 1, drinking >11 alcohol units per week, a marital status of single/divorced/widowed, lower social class, diagnosis of stroke, and type 2 diabetes.

The median worse-ear hearing threshold (referred to as “hearing threshold”) was 27.50 dB HL at timepoint 1, worsening to 38.02 dB HL at timepoint 2. This represents an average decline of 10.5 dB, which is clinically meaningful because decibels are measured on a logarithmic scale. A 10-dB increase corresponds to roughly a doubling in perceived loudness. In practical terms, this means that sounds that were previously audible at the level of a soft whisper (around 25–30 dB) would need to be as loud as a quiet conversation (around 35–40 dB) to be detected. The rate of change in hearing thresholds between the two assessment points exhibited considerable inter-individual variability, with values ranging from a maximum of 6.9 dB per year to a mean of 0.76 dB per year (SD = 1.17). This variation indicates heterogeneity in the progression of hearing decline across participants.

This study investigated the longitudinal associations between hearing threshold and subsequent cognitive function, as well as between hearing threshold and social isolation, in community-dwelling older adults. Over a 10-year period, the cohort’s average hearing threshold declined by 10.5 dB. Contrary to previous research (3, 22, 23), our analysis identified no statistically significant associations between hearing loss and cognitive performance or between hearing loss and social isolation. However, the relatively small sample size of the study means we were likely underpowered to detect small-to-moderate effects. Future studies need to be of sufficient sample sizes to ensure these small effect sizes do not go undetected. Although not statistically significant, the associations were inverse, with effect estimates and confidence intervals aligning with trends reported in prior studies (24, 25).

Given the sample size, it is important to interpret these results in the context of the confidence intervals, rather than relying solely on statistical significance. For cognition, the 95% CI for the effect of a doubling of hearing threshold (−2.07–0.03) suggests that worse hearing could be associated with up to a two-point lower MMSE score, which would be clinically important, though the interval also includes no effect. For social isolation, the confidence intervals span both modest decreases and increases in social isolation score, indicating that the true association could be small in either direction. While p-values indicate a lack of statistical significance, the confidence intervals for MMSE suggest that clinically meaningful effects cannot be ruled out. Several factors may account for these findings here. A recent cohort study similarly reported no significant effect of hearing loss on cognitive decline after adjustment for age (26), highlighting the importance of controlling for confounding variables and employing outcome measures sensitive to subtle, long-term changes. The observed hearing decline of 10.5 dB over 10 years is relatively modest and may not be clinically meaningful, as declines of 15 dB or more at specific frequencies are typically considered significant in older adults (27, 28). If hearing changes were insufficient to affect daily functioning, null associations with cognition and social isolation would be expected. Furthermore, most participants maintained MMSE scores within the normal range, suggesting minimal cognitive impact from the observed hearing decline.

Our findings diverge from much of the existing literature, and several explanations are plausible. First, our analysis controlled for thirteen potential confounders, including age. Additionally, publication bias may contribute to the predominance of significant findings in the literature. Null results such as ours provide valuable evidence in a field where non-significant findings are under-represented, potentially distorting perceptions of the true relationship between hearing loss and cognition (29, 30).

The use of complete-case analysis introduces potential selection bias. Attrition between baseline and follow-up was substantial, and non-responders were older, had poorer hearing, and exhibited more comorbidities, which were the group most likely to demonstrate associations between hearing loss and cognitive or social outcomes. This “healthier survivor” effect may have attenuated observed associations, warranting cautious interpretation of the null findings. Our models also assumed stability in covariates such as marital status, health conditions, and social class over the 10-year interval. In reality, these factors likely changed and could influence both hearing and cognitive/social outcomes. The inability to account for time-varying confounding represents a key limitation and may have contributed to the null associations observed. Our relatively small sample size compared with previous studies (3, 22) likely reduced statistical power, limiting our ability to detect associations due to underpowering. We used the MMSE to assess cognitive performance; however, despite its widespread use, the MMSE is primarily a screening tool for cognitive impairment and has limited sensitivity to subtle changes associated with normative aging. This limitation is evident in our sample, where 92.9% of participants scored above the commonly used cutoff of 25, suggesting a ceiling effect that may have obscured associations. A more comprehensive psychometric battery would likely provide greater sensitivity to age-related cognitive variation. Similarly, epidemiological measures of social isolation may not capture nuanced patterns as effectively as individual-level observations. Future research should incorporate more sensitive cognitive assessments and consider mediation analyses to explore the role of social isolation in hearing–cognition pathways.

Interpretation of these findings should also consider sample characteristics. While broadly representative of community-dwelling older adults in England and Wales, the sample may not reflect individuals with concurrent hearing loss, cognitive decline, and social isolation. If levels of cognitive impairment and isolation are low, there may be insufficient variability to detect associations. Future studies should adopt inclusive recruitment strategies and longer follow-up intervals to capture a wider range of outcomes and reduce attrition bias (31).

Biological and psychosocial mechanisms remain important considerations. Age-related hearing loss has been shown to affect cognition in a domain-specific manner, with effects varying across dementia subtypes. Animal studies indicate that prolonged moderate hearing loss can impair working and recognition memory, although the severity and pattern of deficits likely depend on the memory domain and compensatory mechanisms (32). Hearing loss and social isolation also share a complex, bidirectional relationship, whereby reduced auditory input can lead to social withdrawal, and isolation may exacerbate cognitive decline. This interplay is further complicated by central auditory processing disorders, which have been linked to neuropsychiatric outcomes, such as late-onset depression (33). These findings highlight the multifactorial nature of hearing-related cognitive decline and the need for integrated approaches addressing both auditory and psychosocial factors.

Interestingly, worse hearing was weakly associated with lower social isolation, contrary to theoretical expectations. Given the wide confidence intervals, this may reflect random variation. Alternatively, cohort-specific factors, such as strong family or community support, may buffer against isolation. These findings underscore the complexity of social dynamics and warrant further investigation. The 10-year interval between assessments may also have obscured more proximal or dynamic associations between hearing and social/cognitive outcomes. Future studies should consider repeated measures and shorter follow-up intervals to capture time-varying processes and potential non-linear effects. Alternative modeling approaches, such as examining changes in hearing over time or exploring threshold effects, may yield different insights.

Although our findings were null, they contribute to ongoing debates regarding the mechanisms linking hearing and cognition (34). The absence of associations in this relatively healthy cohort may suggest a threshold effect, whereby only more severe hearing loss impacts cognition, or that resilience factors buffer against decline. These results align more closely with the shared neuropathology hypothesis than the cascade model, though measurement limitations temper strong conclusions.

This study has several strengths, including the use of objective hearing measures via pure-tone audiometry (35) and a validated instrument (MOS Social Support Survey) to assess social isolation, avoiding reliance on proxy indicators such as marital status or living arrangements. However, limitations include potential selection bias due to attrition, modest hearing changes over time, and the use of outcome measures that may lack sensitivity for detecting subtle changes (36). More frequent follow-ups and comprehensive cognitive batteries would enhance future research. Also, we did not calculate additional effect size indices such as ΔR², which could further clarify the contribution of hearing thresholds beyond covariates; future study should report these alongside regression coefficients to strengthen the interpretation.

In conclusion, although this study did not identify statistically significant associations between hearing loss and cognitive decline or social isolation, the confidence intervals suggest that small but potentially meaningful effects cannot be excluded, and the direction of estimates was consistent with prior research. Our findings emphasize the importance of sensitive outcome measures, robust sample sizes, and inclusive study designs. As populations age, integrated strategies that promote hearing health and social engagement may help support cognitive resilience in older adults.