Purpose: Despite modern therapy, patients with low-grade non-Hodgkin's lymphomas (NHLs) have a median survival of only 7 to 10 years. To determine factors that predict for short survival after relapse and thus to identify candidates for intensive investigational studies including bone marrow transplantation, we have analyzed the combined results of three Eastern Cooperative Oncology Group (ECOG) trials of initial chemotherapy for lymphoma. Patients and Methods: All 466 patients who achieved initial complete response (CR) or partial response (PR) and had a subsequent relapse were evaluated (median follow-up, 12.6 years). Multivariate regression analysis within a training set (two thirds of cases) was verified in the remaining one-third (validation set) of cases. Results: Age younger than 60 years, CR, and response duration were significantly associated with longer survival after relapse. Multivariate analysis developed a predictive model that identified shorter survival in patients ≥ 60 years, regardless of CR or response duration. Patients younger than 60 years with an initial CR of more than 1 year had a median survival of 5.9 years, those with a PR of more than 1 year had a median survival of 4.2 years, and those with a CR or PR of ≤ 1 year, 2.4 years (P < .0001). Even the most favorable group had a 10-fold greater mortality compared with age-adjusted United States population rates. Conclusions: These data suggest that patients with low-grade NHLs with a ≤ 1-year response period have poor survival after relapse and may be candidates for aggressive salvage therapy, including transplantation. Longer initial responses lead to better survival after relapse. Clinical trials seeking to demonstrate an advantage for new treatments including transplantation will require long follow-up and comparison to control populations for meaningful analysis.