Graft-versus-host disease (GVHD) is the primary limitation of allogeneic hematopoietic cell transplantation (HCT). Although we have validated a panel of 6 GVHD plasma biomarkers with diagnostic and prognostic value, current blood tests do not predict a patient's response to therapy. We compared plasma from responders (R) and non-responders (NR) using proteomics as a discovery engine, and high-throughput ELISA as validation. Five proteins significantly elevated in NR and 6 validated GVHD markers were measured at therapy initiation in 381 patients with acute GVHD and treated with steroids. In multivariate analysis including the clinical GVHD grade (the strongest clinical predictor), suppressor of tumorigenicity 2 (ST2), the IL33 receptor, significantly predicted response to GVHD therapy. Then, we compared ST2 and grade to predict mortality 180 days post-therapy, we found that patients with high grade and low ST2 had a good prognosis while patients with low grade and high ST2 had a poorer prognosis, suggesting that ST2 provides important prognostic information above the clinical grade. We then analyzed the ability of ST2 measured 14 days post-transplant to predict the occurrence of GVHD and mortality. Patients with high ST2 were 1.4 times more likely to develop GVHD by day 100 post-transplant and 3.2 times more likely to die 180 days post-transplant. Therefore, measurement of ST2 significantly enhances the accuracy of outcome prediction in allogeneic HCT recipients.