One hundred and ninety-three children with T-cell acute lymphocytic leukemia (T-ALL) whose leukemia cells were E-rosette positive were treated on a Pediatric Oncology Group study (1979-1986) designed specifically for patients with T-ALL. The results of modified LSA2L2 therapy with or without intensified intrathecal chemotherapy and cranial irradiation (radiotherapy) were compared with those obtained using a simpler multi-agent protocol which included radiotherapy (T-cell 2). The complete remission (approximately 90%) and 3-year event-free survival rates (approximately 40%) were similar in the three treatment groups. However, the pattern of extramedullary relapse varied according to specific treatment regimen. Patients who received LSA2L2 therapy with less intensive intrathecal chemotherapy and no radiotherapy had a central nervous system (CNS) relapse rate (i.e. isolated CNS +/- other site) of over 20%, compared to only 10% for patients receiving the same systemic chemotherapy with intensified intrathecal therapy and radiotherapy, and less than 5% for those receiving T-cell 2 therapy. In contrast, males receiving T-cell 2 therapy had a testicular relapse rate of greater than 20% compared to less than 10% for patients receiving either regimen (i.e. +/- intensified intrathecal chemotherapy and radiotherapy) of modified LSA2L2 therapy. We conclude that, in the context of these therapies, central nervous system irradiation plus intensive triple (hydrocortisone, methotrexate, cytarabine) intrathecal chemotherapy is more effective than CNS preventative therapy comprised of intrathecal low-dose methotrexate only, and that the more complex multi-agent chemotherapy used in the modified LSA2L2 regimens appeared to be more effective in prevention of testicular leukemia, indicating that the effectiveness of sanctuary site treatment was therapy-specific.