Symptom-level and network analyses of ketamine treatment in treatment-resistant depression: A pooled analysis of two...

We examined symptom-level treatment effects and exploratory symptom networks after ketamine in treatment-resistant depression.

This post hoc analysis pooled 132 participants assigned to ketamine 0.5 mg/kg or protocol-specific control (66 each) from saline-controlled 2017 and midazolam-controlled 2023 trials. Baseline- and trial-adjusted models assessed Montgomery-Åsberg Depression Rating Scale (MADRS) items and four literature-derived domains with false discovery rate (FDR) correction. Sensitivity analyses included trial-specific models and mixed models for repeated measures. Ketamine-arm Gaussian graphical models compared baseline with day 2. Treatment-node Network Intervention Analysis (NIA) estimated conditional associations; item-edge reporting required full-sample selection and at least 50% bootstrap selection.

At 240 min, 9/10 items met the FDR criterion. Adjusted control-minus-ketamine differences were 1.09 (95% confidence interval 0.65-1.54) for apparent sadness, 1.09 (0.64-1.54) for pessimistic thoughts, and 1.10 (0.68-1.52) for suicidal thoughts. All four domains met the criterion at 240 min and day 2. Global strength increased from 3.38 to 4.47 (median permutation p = 0.024), with no detected structure difference (p = 0.433); baseline edge stability was limited. NIA retained nine edges at 50%, four at 60%, and two at 75%, the latter involving apparent sadness at days 3 and 7.

Ketamine showed broad early symptom improvement. Exploratory network associations were sensitive to reporting thresholds and did not establish symptom-specific mechanisms.

UMIN000016985, UMIN000033916 and UMIN000033760.