Patients with recurrent or refractory low-grade non-Hodgkin's lymphoma (NHL) are increasingly treated with myeloablative therapy and autologous stem cell transplantation. However, allogeneic bone marrow transplantation (BMT) is only sporadically performed in such patients. Therefore, we wish to compare treatment results of patients with recurrent or refractory low-grade NHL who underwent allogeneic BMT with those who underwent autologous BMT in our center. Twenty-eight patients were studied. The patients had received 2 to 5 lines of conventional chemotherapy before the BMT procedure. Eighteen patients, all with chemotherapy-sensitive disease at the time of transplantation, underwent autologous BMT and 10 patients, of whom 7 with chemotherapy-resistant disease at the time of transplantation, underwent allogeneic BMT. Furthermore, all allogeneic BMT patients had overt lymphoma infiltration of the BM at the time of transplantation. The conditioning regimen consisted of cyclophosphamide plus total body irradiation in all 28 patients. All allogeneic BMT patients achieved complete remission, 3 patients had a treatment-related death, and 7 patients are alive and disease-free with a median follow-up of 41 months. In contrast, none of the autologous BMT patients died of transplant-related complications. However, despite the fact that all autologous BMT patients had chemotherapy-sensitive disease and partial remission was converted to complete remission by the BMT procedure in 67% of them, only 3 of 18 patients are alive and disease-free. The probability of relapse or disease-progression among allogeneic BMT patients was 0% compared with 83% for autologous BMT patients (P = .002). Progression-free survival rates 2 years after BMT were 68% for allogeneic BMT patients and 22% for autologous BMT patients (P = .049). Although the numbers of patients are small, this study suggests that allogeneic BMT offers a better chance for cure than autologous BMT for patients with poor-prognosis low-grade lymphoma, and the difference in relapse or disease progression is strongly suggestive for the existence of a graft-versus-low-grade lymphoma effect.