Supplementary Material for: Anxiety trajectories during electroconvulsive therapy and their role in predicting depression treatment response: towards personalized therapy

Introduction: Current predictors of electroconvulsive therapy (ECT) effects rely on static measures. The prognostic value of anxiety evolution during ECT remains unestablished, which impedes personalized treatment. This study aims to investigate the dynamic evolution of anxiety during ECT and its predictive role in treatment effects. Methods: We collected 1,053 data points from 117 patients with depression who were undergoing ECT, measuring both the Hamilton Anxiety Rating Scale (HAMA) and the 17-item Hamilton Depression Rating Scale (HAMD-17) at baseline and at each of the first eight ECT sessions. K-means longitudinal (KML) clustering identified anxiety evolution patterns, while linear mixed-effects modeling (LMM) and survival analysis were used to assess associations with treatment outcomes. Results: Three distinct anxiety trajectory phenotypes were identified: Rapid Anxiety Remitters (Cluster A: 47%), Gradual Anxiety Improvers (Cluster B: 41%), and Anxiety Non-Remitters (Cluster C: 12%). LMM analysis demonstrated significantly greater HAMD-17 reduction slopes in Cluster A (β = -1.32, p = 0.003) and Cluster B (β = -1.15, p = 0.007) compared to Cluster C. Kaplan-Meier analysis revealed a hierarchical response advantage (Cluster A > B > C; Log-Rank χ² = 32.15, p < 0.001). Multivariable Cox regression confirmed the superior predictive value of trajectories over baseline HAMA: Rapid Anxiety Remitters showed a 19.77-fold higher response likelihood than Anxiety Non-Remitters (95%CI 8.21-47.63), compared to only a 4% increased probability per HAMA point (hazard ratios = 1.04). Conclusion: Dynamic anxiety trajectories are effective biomarkers for ECT response, offering enhanced predictive value over baseline HAMA scores and enabling personalized treatment strategies for depression.