Presymptomatic central nervous system (CNS) treatment in children with a late isolated first bone marrow (BM) relapse of acute lymphoblastic leukemia (ALL) was based on intermediate-dose systemic and intrathecal (IT) methotrexate (MTX) in the multicenter trial, ALL-REZ BFM 85. Because this was associated with an excess of overt second CNS relapses, cranial radiotherapy (cRT) plus prolonged triple IT therapy with MTX, cytarabine, and prednisone was instituted during the course of the subsequent trial, ALL-REZ BFM 87. Results of children with or without cRT, but otherwise identical chemotherapy, are presented here. Between April 1985 and March 1990, 93 children with their first late isolated BM relapse of ALL were entered on protocols ALL-REZ BFM 85M and ALL-REZ BFM 87. An intensive 6-month phase of multiagent chemotherapy that included 8 courses of systemic MTX (1 g/m2) plus IT MTX was followed by 2 years of conventional maintenance therapy with daily 6-thioguanine and biweekly MTX. Children with bone marrow transplantation excluded, 73 were in complete remission at the end of intensive polychemotherapy, 40 of whom received fractionated cRT plus triple IT therapy during the following 6 months; 11 did not receive cRT but prolonged triple IT; 22 received neither cRT nor prolonged triple IT. Except for a higher percentage of children who had received cRT in front-line protocols (29 of 33 v 20 of 40), the patient groups without or with salvage cRT were comparable. Of 33 children without salvage cRT, 26 relapsed, compared with 21 of 40 who had received cRT (P < .05). The difference was solely attributable to second relapses with CNS involvement (10 of 33 v 1 of 40; P < .01). Estimated 6-year event-free survival rates were .18 for children without cRT and .46 for children with cRT (P < .01). In patients without cRT, no impact of prolonged IT therapy could be shown. The data suggest that second CNS prophylaxis with cRT and prolonged triple IT therapy in children with late isolated BM relapse of ALL is effective in preventing CNS relapses, in reducing the overall relapse rate, and in increasing the overall survival rate.