Baseline effective connectivity predicts response to repetitive transcranial magnetic stimulation in patients with treatment-resistant depression
URL with Digital Object Identifier
Repetitive transcranial magnetic stimulation (rTMS)has become a popular treatment option for treatment-resistant depression (TRD). However, suboptimal response rates highlight the need for improved efficacy through optimisation of treatment protocol and patient selection. We investigate whether the limbic salience network and its connectivity with prefrontal stimulation sites predict immediate and longer-term responsiveness to rTMS. Twenty-seven patients with TRD were randomly allocated to receive 16 sessions of either conventional rTMS or intermittent theta-burst (iTBS)over 4 weeks; delivered using connectivity profiling and neuronavigation to target person-specific dorsolateral prefrontal cortex (DLPFC). At baseline and 3-month follow-up, patients underwent clinical assessment and scanning session, and 1-month clinical follow-up. Resting-state fMRI data were entered into seed-based functional and effective connectivity analyses between right anterior insula (rAI)and DLPFC target, and independent components analysis to extract resting-state networks. Cerebral blood flow (CBF)was also assessed in the rAI. All brain measures were compared between baseline and follow-up, and related to treatment response at 1- and 3-months. Baseline fronto-insular effective connectivity and salience network connectivity were significantly positively correlated, while baseline rAI CBF was negatively correlated, with early (1-month)response to rTMS treatment but not sustained response (3-months), suggesting persistence of therapeutic response is not associated with baseline features. Connectivity or CBF measures did not change between the two time points. We demonstrate that fronto-insular and salience-network interactions can predict early response to rTMS in TRD, suggesting that these network nodes may be key regions toward developing rTMS response biomarkers.