Utilizing stimulation-evoked hemodynamic activity to predict antidepressant response to intermittent theta-burst stimulation in adults with major depression.

BACKGROUND

The instantaneous neural response to prefrontal theta burst stimulation (TBS) may serve as predictive marker for antidepressant treatment success. This study aimed to (1) assess whether baseline theta burst stimulation (TBS)-induced prefrontal hemodynamic responses can predict treatment outcome of four weeks of TBS in adults with major depressive disorder (MDD); (2) assess the test-retest reliability of TBS-induced hemodynamic responses.

METHODS

Forty-four MDD participants were recruited and underwent two consecutive-day concurrent TBS/functional near-infrared spectroscopy (fNIRS) measurements. Participants then received four weeks intermittent TBS (iTBS) treatment. An additional 45 healthy controls (HCs) were recruited for the test-retest reliability analysis. Baseline TBS-induced hemodynamic responses were utilized to classify treatment responders via logistic regression and supervised machine learning. Intraclass correlation coefficients (ICCs) were calculated using a two-way mixed-effects model with absolute agreement to assess the test-retest reliability of TBS-induced hemodynamic responses.

RESULTS

A logistic regression model distinguished responders from non-responders (R2 = 0.617, p < 0.001) and a support-vector machine classifier achieved an accuracy of 82.9% and an AUC of 0.902 in identifying responders. The test-retest reliability of TBS-induced prefrontal hemoglobin responses (Single Measures ICCs) ranged from 0.301 to 0.752, suggesting poor to excellent reliability.

CONCLUSIONS

TBS-induced prefrontal hemodynamic response provides valuable information for predicting antidepressant treatment response, highlighting its potential as predictive imaging marker. The reliability of TBS-induced prefrontal response is comparable to previous neuroimaging marker studies, although better control of external factors is needed to enhance reliability.