Trajectory Modeling and Response Prediction in Transcranial Magnetic Stimulation for Depression

Area of Science:

• Neuroscience
• Psychiatry
• Biostatistics

Background:

• Repetitive transcranial magnetic stimulation (rTMS) is a treatment for depression.
• Predicting patient response to rTMS is crucial for optimizing therapy.
• Current methods for predicting response to rTMS are limited.

Purpose of the Study:

• To compare the predictive accuracy of Latent Class Mixture Modeling (LCMM) and Non-Linear Mixed Effects (NLME) modeling for antidepressant response to rTMS.
• To determine if these models can predict clinically meaningful categorical (non)response, not just continuous symptom scores.

Main Methods:

• A naturalistic sample of 238 patients with treatment-resistant depression receiving rTMS was analyzed.
• LCMM and NLME models were developed to analyze antidepressant response trajectories.
• The predictive performance of LCMM and NLME models was compared.

Main Results:

• LCMM trajectories were influenced by baseline symptom severity, but baseline symptoms alone had low predictive power.
• An optimal nonlinear two-class LCMM model predicted patient response at 4 weeks (AUC = 0.70).
• NLME modeling showed slightly improved predictive performance at 4 weeks (AUC = 0.76), with neither model predicting response earlier.

Conclusions:

• Trajectory modeling approaches like LCMM and NLME demonstrate predictive validity for rTMS response.
• These findings suggest that trajectory modeling could potentially guide future treatment decisions in rTMS therapy.