Prospective Validation That Subgenual Connectivity Predicts Antidepressant Efficacy of Transcranial Magnetic Stimulation Sites

Abstract

Background: The optimal target in the dorsolateral prefrontal cortex for treating depression with repetitive transcranial magnetic stimulation (rTMS) remains unknown. Better efficacy has been associated with stimulation sites that are 1) more anterior and lateral and 2) more functionally connected to the subgenual cingulate. Here we prospectively test whether these factors predict response in individual patients.

Methods: A primary cohort (Boston, n = 25) with medication-refractory depression underwent conventional open-label rTMS to the left dorsolateral prefrontal cortex. A secondary cohort (Michigan, n = 16) underwent 4 weeks of sham followed by open-label rTMS for nonresponders (n = 12). In each patient, the location of the stimulation site was recorded with frameless stereotaxy. Connectivity between each patient's stimulation site and the subgenual cingulate was assessed using resting-state functional connectivity magnetic resonance imaging from a cohort of healthy subjects (n = 1000) and confirmed using connectivity from patients with depression (n = 38).

Results: In our primary cohort, antidepressant efficacy was predicted by stimulation sites that were both more anterolateral (r = .51, p < .01) and more negatively correlated with the subgenual cingulate (r = -.55, p < .005). However, subgenual connectivity was the only independent predictor of response and the only factor to predict response to active (r = -.52, p < .05) but not sham rTMS in our secondary cohort.

Conclusions: This study provides prospective validation that functional connectivity between an individual's rTMS cortical target and the subgenual cingulate predicts antidepressant response. Implications for improving the cortical rTMS target for depression are discussed.

Keywords: Depression; Dorsolateral prefrontal cortex; Resting-state functional connectivity; Subgenual cingulate; TMS; Transcranial magnetic stimulation.

Figure 1. Analysis approach

Each patient’s transcranial magnetic stimulation site was recorded using their MRI and a frameless neuronavigation system (A). Stimulation sites on the scalp surface were projected to the nearest location on the brain surface using an automated algorithm (B). Brain coordinates were transformed into standard atlas space and the volume of stimulated tissue was approximated using an existing TMS model (C, green). Functional connectivity between the stimulation site (C, green) and the subgenual cingulate (C, red) was assessed using fMRI data from a large normative cohort of 1000 subjects. Representative time courses from a single subject are shown (D).

Figure 2. Anatomical location of the TMS site predicts antidepressant response

The location of the stimulation site in standard atlas space is shown for each patient in our primary cohort (Boston, N = 25) including examples of a responder (green) and a non-responder (red) (A). Clinical improvement (% change in the Beck Depression Inventory) was predicted based anatomical coordinates of each patient’s stimulation site and a previously published equation (16) (B). There was a significant correlation between measured and predicted clinical improvement (r = .51, p < .01).

Figure 3. Functional connectivity between the TMS site and the subgenual cingulate is an independent predictor of antidepressant response

The stimulation sites for an example responder (green) and non-responder (red) are shown overlaid on a map of functional connectivity with the subgenual cingulate from 1000 healthy subjects, masked to highlight the DLPFC (A). Functional connectivity between each patient’s stimulation site and subgenual cingulate was a significant predictor of antidepressant response (B, r = −.55, p < .005). Subgenual connectivity was a significant independent predictor of antidepressant response in multivariate analyses of baseline clinical variables (C) or all variables with signs of predictive utility, including the coordinate-based Herbsman equation (D).

Figure 4. Subgenual connectivity predicts response to active but not sham stimulation

The anatomical location of the stimulation site is shown for each patient in our secondary cohort (A, N = 16), all of whom received sham stimulation. Subgenual connectivity did not predict sham rTMS response (B, r = .01, p(1) = .49). A subset of these patients went on to get active stimulation at the same stimulation sites (C, N = 12). Subgenual connectivity was a predictor of clinical response to active stimulation, matching results from our primary cohort (D, r = −.52, p(1) < 0.05).