The best treatment approach for children with B-precursor acute lymphoblastic leukemia (ALL) in second clinical remission (CR) after a marrow relapse is controversial. To address this question, we compared outcomes in 188 patients enrolled in chemotherapy trials and 186 HLA-matched sibling transplants, treated between 1991 and 1997. Groups were similar except that chemotherapy recipients were younger (median age, 5 versus 8 years) and less likely to have combined marrow and extramedullary relapse (19% versus 30%). To adjust for time-to-transplant bias, treatment outcomes were compared using left-truncated Cox regression models. The relative efficacy of chemotherapy and transplantation depended on time from diagnosis to first relapse and the transplant conditioning regimen used. For children with early first relapse (< 36 months), risk of a second relapse was significantly lower after total body irradiation (TBI)-containing transplant regimens (relative risk [RR], 0.49; 95% confidence interval [CI] 0.33-0.71, P < .001) than chemotherapy regimens. In contrast, for children with a late first relapse (> or = 36 months), risks of second relapse were similar after TBI-containing regimens and chemotherapy (RR, 0.92; 95% CI, 0.49-1.70, P = .78). These data support HLA-matched sibling donor transplantation using a TBI-containing regimen in second CR for children with ALL and early relapse.