This study compares allogeneic BMT with conventional chemotherapy for childhood ALL in second remission. Seventy-five children were transplanted between July 1981 and December 1995. For each patient two control patients matching the following criteria were selected from the Nordic database of ALL: (1) time of diagnosis, (2) T vs. non-T ALL, (3) site of relapse, (4) initial risk group, (5) sex and (6) relapse < or > or =6 months after cessation of therapy. The minimal time of follow-up was 24 months. Mortality rate in CR2, leukemic relapse rate and the proportion in continued second remission were 16/75 (21%), 22/75 (29%) and 37/75 (50%), respectively. P2.-EFS for the BMT group was significantly better than that for the control group (0.40 vs. 0.23, P = 0.02). Children transplanted for bone marrow relapses in particular had a higher P2.-EFS (0.35 vs. 0.15 for the control group, P<0.01). Also, children grafted for early BM relapses had a higher P2.-EFS (0.32 vs. 0.11 for the control group P = 0.01). The outcome was similar when children were transplanted after early or late relapse. Also, there was no difference in outcome between the BMT and the chemotherapy group for children with late relapses. We conclude that allogeneic BMT with an HLA-identical sibling donor or other family donor should be performed in children relapsing in bone marrow during therapy or within 6 months of discontinuing therapy.