To better define the role of rituximab in salvage and high-dose therapy (HDT) for relapsed or refractory non-Hodgkin's lymphoma (NHL), patients treated before the implementation of rituximab in salvage and HDT (n = 57, control group) were compared with patients with rituximab included in this procedure (n = 36, study group). All patients had been antibody-naive at this point, and analyses were performed separately for 22 and 31 patients with aggressive, and 14 and 26 patients with indolent NHL, respectively. All patients received two courses of salvage therapy, predominantly dexamethasone, BCNU, etoposide, cytosine arabinoside, melphalan. Conditioning regimens included BCNU, etoposide, cytosine arabinoside, melphalan; BCNU, etoposide, cytosine arabinoside, cyclophosphamide or total body irradiation and cyclophosphamide, with rituximab added for patients in the study group. Despite the absence of differences in stem cell collection, haematopoietic recovery was delayed in patients with aggressive NHL treated in the study group: median days to absolute neutrophil count more than 0.5 x 10(9)/l, 11 vs 10 (p = 0.01), and platelets more than 20 x 10(9)/l, 14 vs 11 (p = 0.0005), with an increased requirement for platelet transfusions. No similar observations were made in indolent lymphoma patients. Remission rates were superior for patients with aggressive NHL in the study group. With a median follow-up of 7.25 and 4.5 years, this resulted in an improvement in OS at 4.5 years: 67 vs 45% (95% confidence interval, 47-87% vs 28-64%; p = 0.0468). For patients with indolent lymphoma, no comparable benefit was detectable. Our data support the use of rituximab in HDT for patients with aggressive NHL. For patients with indolent NHL, only longer follow-up and/or randomized trials may help to fully determine the impact of rituximab on the outcome after HDT.