What question did this study set out to answer?

To assess whether early symptom improvement during TMS treatment can predict final treatment outcomes.

June 18, 2026Journal of Affective Disorders Reports1 citationsOpen Access

Predicting Final TMS Outcomes From Symptom Trajectory During Treatment

YZYalun ZhangBoston University MRMahdi RazafshaHarvard University LSLera SterinaBoston University

Key Points

To assess whether early symptom improvement during TMS treatment can predict final treatment outcomes.
Retrospective analysis of clinical data from 458 patients treated with TMS targeting the dorsolateral prefrontal cortex.
Employed logistic regression to predict responder status and multiple regression to assess prediction of final improvement.
Assessed symptom improvement at multiple time points using the Quick Inventory for Depressive Symptomatology.
Symptom improvement during TMS significantly predicted higher probability of final response.
Model discrimination improved from treatment day 10 to day 20, with no further increase observed by day 30.
Correlation coefficients indicated strong predictive power of early improvements for final QIDS change, reaching R=0.73 by treatment day 30.

Abstract

Background Symptom improvement during TMS treatment may predict treatment response, but it remains unclear if this early change is clinically useful. Objective To evaluate the predictive value of symptom improvement during TMS treatment for outcomes in patients receiving TMS for major depressive disorder who have not yet achieved response criteria. Methods We retrospectively analyzed clinical data from 458 patients treated with Brainsway H1 coil targeting dorsolateral prefrontal cortex. Quick Inventory for Depressive Symptomatology (QIDS) was collected before treatment 10, 20, 30, and 36. Analyses included logistic regression predicting responder status by treatment end from symptom improvement during TMS treatment, linear regression predicting QIDS change at treatment end from QIDS change during treatment; and multiple regression to test whether individual symptom improvement predicts final improvement. Results Symptom improvement during TMS treatment predicted higher probability of final response, with model discrimination increasing from treatment 10 (T10) to treatment 20 (T20), but no further increase at treatment 30 (T30). In a leave-one-out cross-validation, there was strong correlation between early and final predicted percent QIDS change at treatment 36 based on early improvement at T10 (R=0.42, p<10 -15 ), T20 (R=0.59, p<10 -15 ), and T30 (R=0.73, p<10 -15 ). Based on these data, we report a formula and a free web-app to predict final treatment response based on early response. No individual QIDS item significantly predicted final improvement after false discovery rate correction. Conclusions Symptom improvement during TMS treatment, measured as percent improvement, raw change, or empirical response rate—provides robust, clinically interpretable prediction of final TMS response in patients at our center who have not yet met early response criteria. We provide a free web app for automatically predicting final response based on early improvement (moiraitms.streamlit.app).

问 AI

Bookmark

View Full Paper