Allogeneic stem cell transplantation (SCT) is frequently considered as treatment for relapsed childhood acute lymphoblastic leukemia (ALL). For patients without a matched sibling donor, SCT from unrelated donors (UD-SCT) has been increasingly performed during the past years. However, UD-SCT-related mortality and morbidity is still considerable, and the question remains as to which patients are at such high risk of recurrence that UD-SCT is indicated and, conversely, which patients do not require transplantation for long-term disease control. A matched-pair analysis was performed among patients treated according to Acute Lymphoblastic Leukemia Relapse Berlin-Frankfurt-Münster (ALL-REZ BFM) Study Group protocols after first relapse with chemotherapy or UD-SCT. Altogether 81 pairs were identified that could be matched exactly for site of relapse and immunophenotype, and as closely as possible for duration of first remission, age, diagnosis date, and peripheral blast cell count at relapse. No significant difference in the probability of event-free survival (pEFS) between UD-SCT and chemotherapy existed regarding 28 pairs with an intermediate prognosis (0.39 +/- 0.10 vs 0.49 +/- 0.11, P =.105), whereas the pEFS was significantly different in the 53 pairs with a poor prognosis (0.44 +/- 0.07 vs 0.00 +/- 0.00, P <.001). The major reasons of treatment failure among patients who underwent UD-SCT were therapy-related death (TRD; 24/81) and relapses (20/81). In contrast, TRD rarely occurred in patients treated with chemotherapy alone (3/81), but relapse was much more common (62/81). In conclusion, UD-SCT provides better event-free survival for children with high-risk relapsed ALL. However, there is no clear advantage of UD-SCT in patients with intermediate prognosis.