Rituximab (RTX) is an effective treatment for children with steroid-resistant immune thrombocytopenia (ITP), but reliable biomarkers to predict its response early are lacking. We analysed urine samples from 37 steroid-resistant ITP patients-17 RTX responders (RTX-R) and 20 RTX non-responders (RTX-NR)-along with 40 healthy controls using a discovery-validation proteomics workflow. In the discovery cohort, we identified 78 differential proteins (DPs) before treatment and 67 DPs after treatment using the data-independent acquisition (DIA) approach. The RTX-R group was associated with humoral immunity and complement activation before treatment, while the RTX-NR group showed stronger connections to cellular immunity and glycolysis. The Mfuzz analysis indicated that RTX worked by reducing the B-cell receptor pathway and lipid metabolism while enhancing platelet activation. We validated two proteins (KLK6 and FUBP1) as candidate biomarkers for predicting RTX response through parallel reaction monitoring (PRM) technology, achieving an area under the curve (AUC) of 0.98, with a sensitivity of 1.00 and specificity of 0.85. A monitoring model with HK1, FUBP1, LAIR2, CSTA and RPL12 differentiated RTX-R from RTX-NR after treatment, yielding an AUC of 1.00, sensitivity of 0.92 and specificity of 0.95. Overall, these findings enhance our understanding of RTX mechanisms and support personalized therapy development for ITP.