Serum inflammatory marker predictors of psilocybin response in treatment resistant depression

Abstract

Using serum samples collected from the PsiDeR trial—a randomised, double‑blind, placebo‑controlled study of a single 25 mg dose of psilocybin for treatment‑resistant depression (protocol: BMJ Open 2021;11:e056091)—we will investigate whether inflammatory marker trajectories and baseline inflammatory status are associated with antidepressant response at week 3 (endpoint). The inflammatory markers assessed are IFN‑γ, IL‑1β, IL‑2, IL‑4, IL‑6, IL‑8, IL‑10, IL‑12p70, IL‑13, TNF‑α, uPAR, and GDF‑15. All markers will be log‑transformed prior to analysis to address expected positive skew and outliers identified as +/- 3 SD form the mean. A broad confounder screen will be performed to test influences of for example, age, sex, ethnicity, smoking, depression severity at baseline, any medications, and this will inform covariate selection. Primary objective (H1): baseline inflammatory markers predicting response in the psilocybin arm. The primary objective is to test whether baseline inflammatory markers predict antidepressant response among participants receiving psilocybin (n=26). Analyses will be restricted to the psilocybin arm. For each marker, we will fit a logistic regression model with week‑3 MADRS response/non-response status (responders being those participants who had a >= 50% reduction in their baseline MADRS) as the dependent variable and baseline cytokine level as the primary predictor, adjusting for relevant covariates based on the confounder screen. Models will be run separately for each marker. False discovery rate (Benjamini–Hochberg) correction will be applied. Bonferroni correction was considered but deemed overly conservative given expected intercorrelations among cytokines. Exploratory analyses will additionally fit models including both treatment arms with cytokine × treatment interaction terms to assess whether baseline markers differentially predict outcomes across psilocybin and placebo. These analyses are exploratory and will not contribute to confirmatory inference. Additional exploratory analysis will investigate association of baseline cytokine with outcome MADRS as a continuous measure, using linear regression. Secondary objective (H2): inflammatory marker trajectories across treatment arms. To test whether inflammatory trajectories differ between psilocybin and placebo, we will fit linear mixed‑effects models (LMMs) for each cytokine. Cytokine level will be the dependent variable, with fixed effects for time (categorical: V2/baseline, V4/day 1, V5/week 1, V6/week 3), treatment group, and their interaction, and random intercepts for participant, adjusting for relevant covariates based on the confounder screen. Baseline (V2) will serve as the reference level. A compound‑symmetry covariance structure will be specified. Multiple‑testing procedures will mirror those used for H1. The confirmatory test for H1 is the omnibus time × treatment interaction, assessing whether cytokine trajectories differ between groups. Planned contrasts (via emmeans) will explore group differences at day 1 (H1a) and week 3 (H1b). Week 1 (V5) will be included to maximise data use but will not be interpreted as a planned contrast due to limited sample size (psilocybin n=9, placebo n=10). As a sensitivity test, paired t-tests of week 3 cytokine level vs baseline cytokine level will be carried out for psilocybin arm and placebo arm separately. Exploratory objective (H3): magnitude of cytokine change correlating with response. We will also examine whether responder/non-responder status is associated with the change in inflammatory marker levels from baseline to week 3 (V6 − V2) in the psilocybin arm (n=23). Linear regression models will be fitted with week‑3 minus baseline cytokine as the outcome and response status as the predictor, adjusting for covariates based on confounder screen. Multiple‑testing procedures will mirror those used for H1.