Distinct changes in the morphology of cortical and subcortical grey matter associated with age-related hearing loss and tinnitus in the UK Biobank participants

Abstract

Prevalence of both hearing loss and tinnitus increases with age. However, neuroimaging studies of both conditions report inconsistent changes in brain morphology likely due to small sample size and variable methodology. Structural and functional neuroimaging studies in hearing loss and tinnitus have revealed distinct neural correlates, and further replication is needed to confirm these findings. This study aims to investigate the effects of hearing loss and tinnitus on the brain morphology in a well-powered sample. We utilized self-reported hearing difficulty and tinnitus in participants with magnetic resonance imaging (MRI) in the UK Biobank cohort. Control participants without hearing difficulty and tinnitus were age and sex matched leading to total sample sizes of 13 074 and 6242 for self-reported hearing difficulty and tinnitus, respectively. We utilized the rich UK Biobank dataset (i) to reveal these brain changes in a well-powered large study of hearing loss and tinnitus, (ii) to document the effect of confounding factors on these associations, (iii) to discriminate the effects of tinnitus versus hearing difficulty on the brain and (iv) to estimate the brain-age gap in hearing difficulty and tinnitus subjects compared with controls. Hearing difficulty is significantly associated with smaller grey matter volumes exclusively in the bilateral transverse temporal regions, whereas tinnitus is associated with larger volumes of bilateral hippocampi and thalami when compared with the control group. Furthermore, correcting for confounders (i.e. diabetes, cardiovascular disease, age, sex, smoking, alcohol consumption and Townsend deprivation index) during statistical analysis helped to better delineate the impact of hearing status on brain structural changes. The brain-age gap analysis showed that participants with tinnitus appeared to have significantly younger brains than controls, whereas participants with hearing difficulty did not differ significantly from the control group. Altogether, our results confirmed previous findings and suggest the enlargement of bilateral thalami as the main effect in people with tinnitus. We also established that there are independent and distinct brain pathologies between hearing difficulty and tinnitus. Therefore, the self-reported measure is a reasonable approach to assess the hearing loss and tinnitus pathologies.

Keywords: Heschl’s gyrus; age-related hearing loss; neuroimaging; thalamus; tinnitus.

Graphical Abstract

Figure 1

Flow chart describing case-control assignments and sample sizes for the hearing difficulty and tinnitus phenotypes and brain-age estimation study.

Figure 2

The association between hearing difficulty and brain structure detected in structural T1-weighted MRI. This association was investigated in the multiple linear regression models (N_cases = 6537 and N_controls = 6537) with (A) Regression Model 1 without additional covariates and (B) Regression Model 2 with additional covariates added. (B) Additional row showing the inflated cortex representation to better highlight the Heschl’s gyri. Colours relate to the T-value of the effect sizes from both multiple linear regression models collectively ranging from −5.84 to 5.84. Negative and positive effect sizes indicate cortical/subcortical thinning and thickening, respectively, in hearing difficulty cases compared with controls. FDR, false discovery rate.

Figure 3

The association between tinnitus and brain structure detected in structural T1-weighted MRI. This association was investigated in the multiple linear regression models (N_cases = 3121 and N_controls = 3121) with (A) Regression Model 1 without additional covariate and (B) Regression Model 2 with additional covariates. Colours relate to the T-value of the effect sizes from both multiple linear regression models collectively ranging from −6.89 to 6.89. Negative and positive effect sizes indicate cortical/subcortical thinning and thickening respectively in tinnitus cases compared with controls. FDR, false discovery rate.

Figure 4

Plot comparing T-values from hearing difficulty and tinnitus linear regression (Model 2). Analyses adjusted for age, sex, intracranial volume and further health-related covariates. Each data point represents IDP in two different conditions. The straight line represents the expected association between T-values for hearing difficulty (N_total = 13 074) and tinnitus (N_total = 6242) from mixed effect linear Model 2 after accounting for sample size differences (reference line: intercept = 0, slope = sqrt(13 047/6242). FDR, false discovery rate; P_FDR, FDR-corrected P-value; IDP, imaging-derived phenotype; LH, left hemisphere; RH, right hemisphere.

Figure 5

Boxplots on the comparisons of pBAG between control, hearing difficulty and tinnitus groups. The boxplot depicting the pBAG (predicted brain age—chronological age) for each group (control, hearing difficulty and tinnitus). Welch’s two-sample T-tests were used between hearing difficulty and tinnitus with control group separately. Number of subjects in control (N = 15 112), hearing difficulty with and without tinnitus (N = 8741) and tinnitus with and without hearing difficulty (N = 3839) groups. pBAG, predicted brain-age gap.