Insular subregional dysfunction in panic disorder: A multimodal MRI study

The spatial organization of insular abnormalities across connectional-architecture-defined subregions in panic disorder (PD) remains unclear. We examined structural and resting-state functional measures using the Brainnetome Atlas.

Forty-nine medication-naïve patients with PD and Fifty healthy controls underwent structural and resting-state MRI. Gray matter volume (GMV), amplitude of low-frequency fluctuation (ALFF), and regional homogeneity (ReHo) were assessed across six insular subregions. Four prespecified functional connections were examined, together with exploratory Granger causality analysis between dIg and S1 and an AAL3 whole-insula comparison.

dIg-S1 functional connectivity was reduced in PD (β = -0.148, 95% CI [-0.248, -0.048], pFDR = 0.017). ALFF showed a Group × Subregion interaction (F(5,578) = 4.162, p = 0.001). Only the ventral agranular insula (vIa) showed significantly higher ALFF after false discovery rate (FDR) correction (pFDR = 0.016), with a larger group effect than other subregions. The interaction persisted after temporal signal-to-noise ratio (tSNR) adjustment (F(5,577) = 2.582, p = 0.025), but the vIa-specific effect was nonsignificant in a separate vIa-only tSNR-adjusted model (p = 0.194). ReHo was directionally lower across all six subregions, but no contrast survived FDR correction. No robust subregional GMV differences or significant group differences in exploratory Granger causality were detected. Adjusted AAL3 whole-insula differences were nonsignificant.

PD showed measure-dependent insular functional alterations, including reduced dIg-S1 coupling and vIa-predominant ALFF differences, with the latter sensitive to signal quality and spatial smoothing.