The assumption that anxiety disorders uniformly accelerate brain aging may need revision. For the millions of adults living with generalized anxiety disorder, this distinction matters: are their brains aging faster across the board, or is something more nuanced — and potentially more actionable — occurring at the neurological level?
Using a convolutional neural network trained on structural MRI data from over 3,500 healthy controls and validated against nearly 6,150 individuals — one of the largest neuroimaging datasets assembled for anxiety research through the ENIGMA-Anxiety GAD Working Group — investigators calculated predicted brain age difference (PAD) by subtracting chronological age from AI-estimated brain age. The model achieved a mean absolute error of roughly 2.95 years, competitive with leading image-based brain age benchmarks. Crucially, individuals with GAD did not show systematically older-appearing brains than healthy controls. Instead, the GAD group demonstrated significantly greater *variability* in PAD, particularly among those over 25. A subset analysis also linked symptom severity to PAD independent of medication use and comorbid diagnoses.
This heterogeneity finding reframes how clinicians and researchers should interpret brain aging in psychiatric populations. Prior studies linking anxiety to accelerated aging often used smaller samples or cruder neuroimaging methods, possibly conflating variance with mean shift. The current work suggests GAD may not uniformly age the brain but may create divergent aging trajectories — some individuals aging faster, others slower — which would be invisible in group-mean analyses. This has practical implications: brain age variability could serve as a biomarker for identifying GAD subgroups with distinct prognoses or treatment responses. Limitations include the cross-sectional design, which cannot establish causality or track individual trajectories over time, and potential site-level variation across the multi-cohort ENIGMA dataset. Replication with longitudinal data will be essential before clinical translation. Overall, this is an incremental but methodologically rigorous contribution that usefully complicates the prevailing narrative.
