Individual differences in hearing decline predict region-specific brain and cognitive changes in healthy aging men

Abstract

Background Cognitive decline in aging is heterogeneous and accelerates after approximately 60 years of age. Hearing decline is a common, potentially modifiable risk factor that may contribute to individual differences in cognitive aging and brain structural change. In addition to its cognitive consequences, hearing loss can impair communication and contribute to reduced social participation, loneliness, and poorer quality of life. However, the extent to which longitudinal changes in hearing predict cognitive performance and regional brain structure remains unclear. Methods We assessed 51 healthy males from a longitudinal birth cohort extensively evaluated with cognitive testing and MRI across the lifespan. At ages 70 and 71 (recruited in 2024), participants completed four-frequency pure-tone audiometry (4PTA), audible contrast threshold (ACT), speech-in-noise tests (HINT), cognitive assessments, and structural MRI scans. All participants had a prior visit at age 63 ± 1 year (recruited in 2015–2018) with the same cognitive test battery. Previously, machine learning algorithm models were used to identify cognitive decline, and participants were clustered into four cognitive trajectories. These predefined clusters were used to examine whether hearing measures (4PTA, ACT, HINT) predicted cognitive outcomes and brain volume. Findings The combination of PTA and HINT together showed the strongest contribution to classifying previously defined cognitive trajectory groups (p = 0.022). An XGBoost model with SHapley Additive exPlanations (SHAP) values identified bilaterally combined temporal lobe volume as key MRI feature influencing ACT, whereas cerebellum and fourth ventricle were key MRI features influencing HINT. Interpretation These findings highlight complex hearing–cognition–brain interactions and support integrating diverse auditory measures into cognitive aging research.