Differential sensitivity of MRI-based brain-age models to chronic noise exposure: an exploratory comparative study

Chronic noise exposure is associated with cardiovascular and mental health outcomes, yet its neurobiological effects remain poorly characterized. MRI-based brain-age models estimate structural brain aging, but their sensitivity to environmental and occupational noise exposure has not been systematically evaluated. We examined associations of environmental and occupational noise exposure, separately and in combination, with brain age gap (BAG) across six MRI-based brain-age prediction models. We analyzed 108 Korean hospital workers who underwent high-resolution three-dimensional T1-weighted magnetic resonance imaging. Residential environmental noise exposure was classified as high when daytime noise was ≥60 dB or nighttime noise was ≥55 dB. Because no participant had occupational noise exposure ≥85 dBA as an 8-h time-weighted average, occupational exposure was categorized as relatively lower or higher using the cohort-specific median. BAG was calculated as predicted brain age minus chronological age using brainageR, XGBoost, ENIGMA, pyment, DenseNet, and TSAN models. Primary analyses used covariate-adjusted linear models for BAG, controlling for age, sex, and intracranial volume. Age-bias-corrected and Z-standardized brain-age advancement (BAA) was used for descriptive joint-exposure visualization and exploratory trend analysis. High environmental noise exposure was associated with a less negative BAG in the TSAN model than low environmental noise exposure (-8.1 vs. -12.1 years; p = 0.0036), indicating a relatively older-appearing brain. Daytime and nighttime environmental noise levels were positively correlated with TSAN-derived BAG (Spearman's rho = 0.253, false discovery rate-adjusted p = 0.008; rho = 0.215, false discovery rate-adjusted p = 0.025, respectively). Occupational noise exposure alone was not significantly associated with BAG in any model. No environmental-noise-by-occupational-noise interaction remained statistically significant after false discovery rate correction. In an exploratory TSAN analysis, each additional high-exposure domain was associated with a 0.27-standard-deviation increase in BAA (95% confidence interval, 0.01 to 0.59; p for trend = 0.048). Environmental noise exposure was associated with relatively older-appearing brain structure in the TSAN model, whereas occupational noise exposure alone was not consistently associated with BAG. Concurrent environmental and relatively higher occupational noise exposure showed a model-specific joint-exposure burden gradient but did not provide formal evidence of synergistic or non-additive effects. Larger longitudinal studies are needed to replicate these findings and clarify the utility of brain-age measures in environmental and occupational health research.