¹National Yang-Ming Chiao-Tung University, Institute of Brain Science, Taipei City, Taiwan, ²Taipei Veterans General Hospital, Department of Psychiatry, Taipei City, Taiwan

Background: Standard repetitive transcranial magnetic stimulation (rTMS) typically requires 20–30 treatment sessions for patients with major depressive disorder (MDD) who have not responded to at least one antidepressant administered at an adequate dose and duration. rTMS for the treatment of MDD was approved in the United States in 2008 and in Taiwan in 2018. Since then, various stimulation protocols, such as unilateral and bilateral dorsolateral prefrontal cortex (DLPFC) stimulation, 10-Hz rTMS, and suprathreshold or subthreshold intermittent theta burst stimulation (iTBS), have demonstrated efficacy in managing pharmacoresistant depression (PRD).
Subthreshold prolonged iTBS (piTBS), consisting of 1,800 pulses over a two-week course, has been shown in two clinical trials to produce antidepressant effects comparable to those of suprathreshold iTBS protocols delivering 600 pulses over 4–6 weeks. A recent real-world study further reported a 57.1% response rate following a 10-session course of subthreshold piTBS.
In clinical practice, predictive models for estimating remission after rTMS intervention are essential for tailoring treatment strategies and improving outcomes in individuals with PRD. Although EEG-based prediction models have demonstrated promising accuracy in estimating remission probability before treatment initiation, there remains a need for prediction models based on clinical characteristics due to their accessibility, lower cost, and independence from specialized equipment.

Aims & Objectives: To create a prediction model for remission based on clinical factors before subthreshold piTBS initiation.

Method: We conducted a retrospective cohort study at a large medical center and recruited patients with non-psychotic PRD who received daily subthreshold piTBS treatment, completing at least 10 sessions under a self-pay model. Based on a review of the literature, five key clinical variables, age, chronicity of the depressive course, history of antidepressant treatment failure, presence of anxious depression, and baseline depression severity, were selected to construct a remission prediction model using logistic regression analysis. Treatment effectiveness was assessed using the 17-item Hamilton Depression Rating Scale (HDRS-17), and remission was defined as an HDRS-17 score of ≤7.

Results: We recruited 119 patients, with a mean age of 39.9 years and a female predominance (64%). The baseline HDRS-17 score was 21, which decreased to 12 following piTBS treatment, yielding a remission rate of 35%. Although there was no biological indeices used, the prediction model accounted for 32% of the variance and demonstrated reasonably good discriminative ability for identifying treatment remitters (AUC = 0.79, 95% CI: 0.70–0.87, p < 0.01), with a sensitivity of 76%, specificity of 69%, and a negative predictive value (NPV) of 85%. The optimal probability threshold based on the Youden index was 0.33.

Discussion & Conclusions: These findings underscore the utility of a prediction model to guide or help clinical decision-making for using subthreshold iTBS treatment in patients with PRD, thereby supporting individualized treatment selection and optimizing therapeutic outcomes.