Corticosteroid resistance is a major obstacle to the management of patients with primary immune thrombocytopenia (ITP), as the factors contributing to the variability in treatment response remain largely unknown. We analyzed a broad set of plasma protein biomarkers from patients who are corticosteroid-sensitive (CSp; n = 30), patients who are corticosteroid resistant (CRp; n = 26), and healthy controls (HCs; n = 25) using a 92-plex immunoassay from Olink Proteomics Technology. A total of 54 inflammation-related proteins demonstrated significant differences among the three groups. Twenty-seven biomarkers showed statistical differences between CSp and CRp. Machine learning-based feature selection identified four potential biomarkers, which were closely related to corticosteroid resistance: CXCL10, IL-1a, glial cell line-derived neurotrophic factor (GDNF), and CCL11. A Nomograph model was developed based on these 4 biomarkers, demonstrating remarkable discriminative ability, with an area under the curve (AUC) of 0.920 (95% confidence interval: 0.830-1.000) in this exploratory cohort. Although requiring external validation in larger studies, the identification of the four potential biomarkers suggested their value in predicting corticosteroid resistance in patients with newly diagnosed ITP, and might guide the initial choice of treatment. (c) 2026 International Society for Experimental Hematology. Published by Elsevier Inc. All rights are reserved, including those for text and data mining, AI training, and similar technologies.