Personalizing neuromodulation for chronic pain: A connectivity-guided trial

In this randomized, double-blind, controlled trial of 8 weeks of repetitive transcranial magnetic stimulation (rTMS) for chronic pain, we compared the classic primary motor cortex (M1) rTMS with a novel target-selection strategy based on pre-therapy cortical connectivity. Guided by principles of homeostatic plasticity, we tested whether stimulating the cortical site with the lowest pre-therapy global connectivity would be more effective than two active comparators: stimulating the site with the highest pre-therapy global connectivity or stimulating M1 independent of connectivity. Before starting rTMS treatment, TMS-evoked EEG potentials were recorded from four cortical targets: M1, the dorsolateral prefrontal cortex, the anterior cingulate cortex, and the posterosuperior insular cortex. For each target, global connectivity was quantified using a distance-weighted, phase-based index (debiased weighted phase lag index, wPLI) derived from pre- and post-TMS-evoked EEG activity, capturing both the magnitude and spatial extent of TMS-induced oscillatory phase locking across cortical regions. Target allocation in the Low- and High-Connectivity groups was based on this global connectivity measure. Ninety patients with chronic pain were randomized to Low-Connectivity, High-Connectivity, or Classic-M1 groups. Treatment consisted of 12 rTMS sessions delivered over 8 weeks to the assigned target. The primary outcome was the proportion of patients achieving [≥] 30% reduction in pain intensity. Secondary outcomes included continuous change in pain intensity, pain interference, sleep, fatigue, mood, quality of life, and patient global impression of change. No between-group differences were observed for primary or secondary outcomes (p > 0.05). In prespecified exploratory analyses, we examined whether pre-therapy local connectivity (within-target wPLI) predicted treatment response. In the Classic-M1 group, lower pre-therapy local M1 connectivity was associated with a greater reduction in pain intensity (r = 0.50, p = 0.005). This association was not observed in the Low- or High-Connectivity groups. A regression model including group-by-connectivity interaction indicated that the relationship between local connectivity and pain reduction differed between the Classic-M1 and High-Connectivity groups (p = 0.038). The results of this clinical trial showed that connectivity-based target allocation using global connectivity did not improve clinical outcomes. However, lower local M1 connectivity was associated with greater pain reduction following Classic-M1 stimulation, suggesting that local M1 connectivity may serve as a potential biomarker of response.