Fig. 1: Study workflow.

a In a double-blind study, patients were randomized to receive a single preoperative dose of 125 mg methylprednisolone (MP, n = 28 patients) or saline placebo (n = 30 patients). Peripheral blood and clinical outcomes data were collected prior to surgery (baseline) and at the indicated time points after surgery. After erythrocyte lysis, peripheral immune cells were barcoded, stained with cell-phenotyping and intracellular cell-signaling antibodies, and analyzed by mass cytometry. Unsupervised bootstrapped clustering of immune cell subsets followed by random forest analysis was performed to identify differential immune cell dynamics in MP vs. control groups. b A non-linear dimensional reduction algorithm (Isomap) showing individual patients’ immunological trajectories after surgery along the innate (X) and adaptive (Y) axes (MP in red, control in black). Snapshots are shown for the 1 h, 6 h, 24 h, 48 h and 2 wk time points (animated trajectories can be found in Supplementary Movie 1). Right panel. Overview of median trajectories for the MP and control groups are shown in red and black.