Functional Connectivity Subtypes During a Positive Mood Induction: Predicting Clinical Response in a Randomized Controlled Trial of Ketamine for Treatment-Resistant Depression

Ketamine has shown promise in rapidly improving symptoms of depression and most notably treatment-resistant depression (TRD). However, given the heterogeneity of TRD, biobehavioral markers of treatment response are necessary for the personalized prescription of intravenous ketamine. Heterogeneity in depression can be manifested in discrete patterns of functional connectivity (FC) in default mode, ventral affective, and cognitive control networks. This study employed a data-driven approach to parse FC during positive mood processing to characterize subgroups of patients with TRD prior to infusion and determine whether these connectivity-based subgroups could predict subsequent antidepressant response to ketamine compared to saline infusion. 152 adult patients with TRD completed a baseline assessment of FC during positive mood processing and were randomly assigned to either ketamine or saline infusion. The assessment utilized Subgroup-Group Iterative Multiple Model Estimation to recover directed connectivity maps and applied Walktrap algorithm to determine data-driven subgroups. Depression severity was assessed pre-and 24-hr postinfusion. Two connectivity-based subgroups were identified: Subgroup A (n 110) and Subgroup B (n 42). We observed that treatment response was moderated by an infusion type by subgroup interaction (p.040). For patients receiving ketamine, subgroup did not predict treatment response (β.326, p.499). However, subgroup predicted response for saline patients. Subgroup B individuals, relative to A, were more likely to be saline responders at 24-hr postinfusion (β 2.146, p.007). Thus, while ketamine improved depressive symptoms uniformly across both subgroups, this heterogeneity was a predictor of pla-cebo response.