Purpose Salvage chemotherapy followed by high-dose therapy and autologous stem-cell transplantation (ASCT) is the standard treatment for relapsed diffuse large B-cell lymphoma (DLBCL). Salvage regimens have never been compared; their efficacy in the rituximab era is unknown. Patients and Methods Patients with CD20⁺ DLBCL in first relapse or who were refractory after first-line therapy were randomly assigned to either rituximab, ifosfamide, etoposide, and carboplatin (R-ICE) or rituximab, dexamethasone, high-dose cytarabine, and cisplatin (R-DHAP). Responding patients received high-dose chemotherapy and ASCT. Results The median age of the 396 patients enrolled (R-ICE, n = 202; R-DHAP, n = 194) was 55 years. Similar response rates were observed after three cycles of R-ICE (63.5%; 95% CI, 56% to 70%) and R-DHAP (62.8%; 95 CI, 55% to 69%). Factors affecting response rates (P < .001) were refractory disease/relapse less than versus more than 12 months after diagnosis (46% v 88%, respectively), International Prognostic Index (IPI) of more than 1 versus 0 to 1 (52% v 71%, respectively), and prior rituximab treatment versus no prior rituximab (51% v 83%, respectively). There was no significant difference between R-ICE and R-DHAP for 3-year event-free survival (EFS) or overall survival. Three-year EFS was affected by prior rituximab treatment versus no rituximab (21% v 47%, respectively), relapse less than versus more than 12 months after diagnosis (20% v 45%, respectively), and IPI of 2 to 3 versus 0 to 1 (18% v 40%, respectively). In the Cox model, these parameters were significant (P < .001). Conclusion In patients who experience relapse more than 12 months after diagnosis, prior rituximab treatment does not affect EFS. Patients with early relapses after rituximab-containing first-line therapy have a poor prognosis, with no difference between the effects of R-ICE and R-DHAP.