The efficacy of allografting in acute lymphoblastic leukemia (ALL) is heavily influenced by remission status at the time of transplant. Using polymerase chain reaction (PCR)-based minimal residual disease (MRD) analysis, we have investigated retrospectively the impact of submicroscopic leukemia on outcome in 64 patients receiving allogeneic bone marrow transplantation (BMT) for childhood ALL. Remission BM specimens were taken 6 to 81 days (median, 23) before transplant. All patients received similar conditioning therapy; 50 received grafts from unrelated donors and 14 from related donors. Nineteen patients were transplanted in first complete remission (CR1) and 45 in second or subsequent CR. MRD was analyzed by PCR of Ig or T-cell receptor delta or gamma rearrangements, electrophoresis, and allele-specific oligoprobing. Samples were rated high-level positive (clonal band evident after electrophoresis; sensitivity 10(-2) to 10(-3)), low-level positive (MRD detected only after oligoprobing; sensitivity 10(-3) to 10(-5)), or negative. Excluding 8 patients transplanted in CR2 for isolated extramedullary relapse (all MRD-), MRD was detected at high level in 12 patients, low level in 11, and was undetectable in 33. Two-year event-free survival for these groups was 0%, 36%, and 73%, respectively (P <.001). Follow-up in patients remaining in continuing remission is 20 to 96 months (median, 35). These results suggest that MRD analysis could be used routinely in this setting. This would allow identification of patients with resistant leukemia (who may benefit from innovative BMT protocols) and of those with more responsive disease (who may be candidates for randomized trials of BMT versus modern intensive relapse chemotherapy).