We have reviewed results of secondary therapy in 427 patients with acute graft-versus-host disease (GVHD) who did not have a durable satisfactory response after primary treatment. At the beginning of secondary treatment, 320 patients (75%) had rash, 252 (59%) had liver dysfunction, and 228 (53%) had gut dysfunction. Secondary treatment was with glucocorticoids (n = 249), cyclosporine (n = 80), antithymocyte globulin (n = 114), or monoclonal antibody (n = 19) either singly (n = 390) or in combination (n = 37). Parameters of GVHD severity were recorded weekly, and responses were determined according to values at the initiation of tertiary treatment or, for patients without such treatment, using values on day 29 of secondary treatment or the last recorded values before death, whichever occurred first. Minimal criteria for improvement or deterioration were defined for each organ, but no attempt was made to define liver or gut outcome if another complication such as venocclusive disease or infectious enteritis was present. Improvement or resolution of GVHD in the respective organ was seen in 45% of patients with skin disease, 25% of patients with evaluable liver disease, and in 35% of patients with evaluable gut disease. Overall complete or partial responses were seen in 40% of patients. The highest complete response rate with secondary therapy (23%) was seen when GVHD recurred during the taper phase of primary glucocorticoid treatment and was managed by increasing the dose of glucocorticoids. Multivariate analyses were performed to identify patient, disease, or treatment factors associated with likelihood of complete response or overall improvement. A similar analysis was performed to identify covariates associated with time to treatment failure (defined as initiation of tertiary therapy or death not due to relapse of malignancy). Severe dysfunction in the skin, liver, and gut at the beginning of treatment was associated both with a decreased likelihood of complete response and an increased treatment failure rate. The times to treatment failure and the proportions of patients in various response categories were similar for primary and secondary treatment, suggesting that the potential efficacy of new immunosuppressive agents for treatment of acute GVHD can be assessed meaningfully in patients who have not responded adequately to initial therapy.